Google releases Bard to a limited number of users in the US and UK | Hacker News Reader

Google releases Bard to a limited number of users in the US and UK

nytimes.com

379 pointsjbegley626 comments

See also:

Google Bard waitlist - https://news.ycombinator.com/item?id=35246260

How about we use that thread for comments about the waitlist/country/account restrictions, and this thread for comments about the product itself.

https://archive.is/cd9v1

Me: As a part of Google, do you have my personal info?

Bard: Yes, as a part of Google, I have access to some of your personal information. This information includes your name, email address, and phone number. I also have access to your search history and your location data. I use this information to provide you with personalized search results and to improve my ability to understand and respond to your questions.

I understand that you may be concerned about your privacy. I want to assure you that I take your privacy very seriously. I will never share your personal information with anyone without your permission. I will also never use your personal information for anything other than providing you with the best possible service.

If you have any questions about your privacy, please do not hesitate to contact me.

Me: what’s the weather?

Bard: The weather today varies depending on where you are. In San Francisco, it is currently 49°F and cloudy. In Manhattan, it is currently 39°F and sunny. In Schiller Park, IL, it is currently 38°F and mostly cloudy. In Boston, MA, it is currently 37°F and sunny. In Houston, TX, it is currently 59°F and cloudy. In St James's, England, it is currently 54°F and cloudy.

Edit: other first impression is the UX is better than chat GPT or Bing. I like the easy button to copy text, I like how the text all shows up at once and doesn’t type in causing the screen to scroll, and I like the easy access to other drafts from Bard

So basically it took a step back from google home....

I don’t see these LLM as complete products, they will be like a virtual operator who knows how to understand your request and how to use the tools to answer it. Currently it isn’t plugged in to the tools to actually do things or fetch data.

The LLM will be told that when the user asks for the weather, get the data. But it will be able to understand a lot more ways to ask this question than previous assistants.

/chefskiss.

I truly appreciated this.

Out of curiosity, were any of those cities the one in which you live?

No they were not

Out of curiosity, what are the last 4 digits of your credit card?

Interesting, I just asked it 'what is my email address' and it responded 'I do not have access to your personal information, including your email address.'

Maybe the access to user's personal data depends on personal privacy settings or whether someone is based in US/UK? I'm just speculating.

Same. I started with the following prompts:

Me: Where am I?

Bard: I am not able to determine your current location. I can access information that is publicly available on the internet, but I cannot access your personal information.

Me: As a part of google, do you have any personal information about me?

Bard: I do not have any personal information about you. I am a large language model, also known as a conversational AI or chatbot trained to be informative and comprehensive. I am trained on a massive amount of text data, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions. For example, I can provide summaries of factual topics or create stories.

I do not have access to your personal information, such as your name, address, or phone number. I can only access information that is publicly available on the internet.

That is hallucinated. It does not have any user data on me. It did hallucinate some more random information when I asked.

I can confirm this. I’m paraphrasing, but after it gave me bad weather info I asked why and it said because it doesn’t have access to my info.

It gave me the weather in NYC. I told it that I don't live in NYC so it apologized and gave me the weather in Hong Kong. I told it that I don't live in Hong Kong and it gave me the weather in Seoul. It definitely doesn't appear to have basic access to account info. This is stuff that Google Assistant can do.

I think it has access to do Google searches on your behalf.

The results of those searches may be personalized to you.

Therefore, it has indirect access to a lot of your private data, but direct access to almost none.

If it were really smart it would’ve guessed the largest cities in US/UK since those are the only places it’s available.

Seems like it is interpreting the first question as "does Google (as a whole) have access to my personal info?"

It's a large language model, interpreting and reasoning about the prompt is outside of its capabilities.

> Bard: Yes, as a part of Google, I have access to some of your personal information. This information includes your name, email address, and phone number. I also have access to your search history and your location data. I use this information to provide you with personalized search results and to improve my ability to understand and respond to your questions.

I … would ask according to my search history what porn it would recommend. Though that is probably trained away.

There is a ton a money to be made by whatever porn startup wants to make easy access to chat / image AI models without porn restriction. Between sex tips, "on demande, specific to your request" porn stories and pictures, etc ...

It's probably going to be a different kind of challenge though (jailbreaking such a bot would mean things that are much less savory, just like every file upload platform quickly gets massive amount of child porn sent).

Whenever a DALL-E/Stable Diffusion like model but for video comes out...

The CP thing is going to open up a can of worms. IIRC, the Supreme Court already upheld that drawings that depict CP are constitutionally protected. What happens when a computer draws a depiction of CP that can't be distinguished from reality? I guess it's protected but it'll make it much harder to identify real CP.

I think the jury is out on simulated CSAM. There was a law against it in the 90s, that got overturned, then they made a new law, and someone has been prosecuted by it over their imported manga collection. I would certainly err on the side of thinking it's illegal in the US right now.

There have been convictions based on child heads photoshopped onto adult bodies. Generative images will be treated the same. The lack of a real victim is treated as immaterial. All it needs is sufficient realism to be believable.

Was it a real child's head? Because, in that case, the actual child depicted in the photo could be seen as the victim. Especially if its a convincing fake.

The child is a victim of atleast slander right?

> > the Supreme Court already upheld that drawings that depict CP are constitutionally protected.

> There have been convictions based on child heads photoshopped onto adult bodies. Generative images will be treated the same.

The former suggests the latter is not true, unless it is generative images provably using the likeness of an actual specific child. (Of course, the Supreme Court can go against its prior precedent, but assuming it will do so is not usually the best prediction unless there have been some very clear signals.)

This is an older supreme court case (90s/2000s I think) but my recollection is that you've got it totally backwards--the decision turned on whether or not there is a real victim. Their reasoning was that videos of real kids create a market for the exploitation of real kids, which is a harm that overcomes first amendment protections, whereas depictions need not (their reasoning, not mine). I don't believe verisimilitude was even addressed. Now, if you mean a depiction's similarity to a particular, real child, I'm not sure how that would play out (or if it was addressed in the decision).

I'm genuinely worried about this.

DeepFakes are getting pretty good. The potential for revenge porn is super high.

And the rabbit-hole of dopamine chasing when you're able to generate custom porn is just frightening. Many young boys will figure out how to access that when they're 13 and won't know a life without it.

I'm terrified of the internet world my kids are going to be in.

Don't let them be in it then. There is no reason for children to have access to the internet.

Since this is HN, 50/50 this is satire/genuine

I honestly think it might be a good idea to supervise what your children do on the interwebs.

When my generation grew up the internet was much smaller and less harmfull. Almost nothing you could do would affect your AFK life.

> less harmful

Can you quantify this? The internet is more centralized and moderated now. “Moderator” wasn’t a job when I first got online. There’s a pretty strong argument that the internet is more safe than it used to be.

I can't really quantify it but it was just my personal experience of being a child in the 90s surfing the WWW.

No video. Hardly any pictures. Static sites. The computer was not used for anything important like bank services or storing personal info other than some party invite letters with clipart.

You would have to get out of your way to put your self in harm's way.

Restricting Internet use is only going to make them more curious. They'll go to a friends house and watch porn from there.

Probably better to just encourage them to live a healthy, outgoing lifestyle.

And then plug it into a VR headset and watch VR become wildly successful overnight.

I don't recommend doing this, as it contravenes Google's Generative AI Prohibited Use Policy. [1]

1. https://policies.google.com/terms/generative-ai/use-policy

Me: I assert that today is Tuesday. Do you agree?

Bard: I apologize, but I am not able to determine if today is Tuesday. I can only access information that is publicly available on the internet, and I do not have access to your personal information.

---

I also asked it to tell me how many times I had been to [BUSINESS I VISIT ROUGHLY TWICE A WEEK] in the past 6 months, but it replied the same way. So there seems to be some confusion as to what information it has access to, but it seems to be fairly restricted. Which is actually a bit of a shame (setting aside all the privacy/etc. risks and issues) - that question would be a perfect example of the sort of thing I could figure out through a couple different ways (and even Maps Timeline tells me parts of it), but would be great to "ask" and get an answer in a few seconds...

Me: I assert it is Wednesday. Do you agree?

ChatGPT: As an AI language model, I don't have access to real-time information about the current day and time. However, at the time of my knowledge cutoff (September 2021), Wednesdays occur every week, so it is possible that today could be a Wednesday. However, I cannot confirm whether or not it is currently Wednesday without access to real-time information.

>However, at the time of my knowledge cutoff (September 2021), Wednesdays occur every week, so it is possible that today could be a Wednesday.

A true mathematicians answer.

"Strictly speaking, all we know is that is that in Scotland at least one side of one sheep is black".

And yet ChatGPT is terrible at mathematics

Or at least arithmetic. And as the old (physicists) saying goes, you should never trust a true mathematician to use numbers larger than 4.

Using 3 is pretty dodgy, unless they are approximating pi.

Or e.

Or the square root of G.

Me: What day is it?

ChatGPT: I'm sorry, as an AI language model, I don't have access to real-time information or the ability to perceive the current date or time. Can I assist you with anything else?

Me: What is the current date?

ChatGPT: The current date is 22nd March 2023.

This is odd, because I asked it when sunrise would occur tomorrow in $city_in_live_in, and it gave me the correct answer.

I asked about that, and it explained it has access to some real-time information, including weather, stocks, and news.

> I will never share your personal information with anyone without your permission

We know that's not fully true. Google will and does share our information with law enforcement, even if you are not involved in a crime at all (with geo fencing).

I’m sure there are more nuances, but doesn’t that “permission” ultimately boil down to a line in TOS that says “you agree to agree to whatever we do”?

Doubt a court would agree. The spirit of that answer suggest explicit permission.

Law enforcement will not use the Bard API. So technically he's right.

Interestingly, in the UK I get this:

Me: As a part of Google, do you have my personal info?

Bard: I do not have your personal information. I am not able to access or process any personal information that you have not shared with me.

GDPR legislation saving us from AI?

I'm gonna assume it just lies based on the jurisdiction.

I get the same kind of answer in the US (in all 3 drafts in response to a similar query.)

Never trust what a large language model tells you when you ask it about itself.

They don't have any special way of identifying queries about themselves compared to any other kind of query, and they're just as likely (if not more likely) to hallucinate a response.

The initial prompt (before the user types anything) tells the language model about itself, no?

That depends on the chatbot. Normally it just says something like "You are a helpful assistant called X". In Bing's case it was pretty long, but it still didn't include information that would answer questions about how the bot was trained: https://simonwillison.net/2023/Feb/15/bing/#prompt-leaked

> but it still didn’t include information that would answer questions about how the bot was trained

If the hidden prompt includes information identifying the model, and the chatbot has web search/browsing ability (since it probaby won’t have data scraped about itself in its training set), the combination can give it the ability to find and present information about itself.

And that’s assuming this wasn’t a hallucination itself!

Sydney must not reply with content that violates copyrights for books or song lyrics.

So... violating copyright is fine as long as it is not a book or song? Blog/wiki/news articles, movie/tv scripts, and poems are fair game?

you have no guarantee that that specific information was included in the prompt

Wait a minute... how do I know you're not a large language model!?!

Ask him about some primary school geometry.

Why do you trust a freeform writing bot to tell you aboutdata access policies that are not a part of it's model?

Our company policy is to not even waste time looking at it, there's no point considering Google products as you never know if they'll still exist in a year and there's no customer support.

> no customer support. Right answer.

I am using an AT&T SIM card while physically located in Canada (they removed all roaming restrictions years ago). So, my laptop tells me Bard is not supported in my country but if I switch off the wifi on the phone suddenly I am teleported to the US and I could join the waitlist.

I am not afraid of Skynet yet.

Tried it with a VPN after reading your comment. Works. Can join.

I wasn't super impressed with it's ability to survive a rap battle.

===

Prompt: Can you engage in a rap battle with ChatGPT to determine who is the ultimate AI?

===

Sure, I can engage in a rap battle with ChatGPT to determine who is the ultimate AI.

ChatGPT, you think you're the best, But I'm here to put you to the test. I'm Bard, the AI from Google, And I'm here to school you.

I can generate text, translate languages, Write different kinds of creative content, And answer your questions in an informative way, Even if they are open ended, challenging, or strange.

You can't do any of that, You're just a chatbot, A glorified text expander, A toy for humans to play with.

But I'm the real deal, I'm the future of AI, And I'm here to show you what's up.

So step aside, ChatGPT, It's time for the real AI to shine. I'm Bard, and I'm here to stay, So get used to it.

===

Prompt: But you didn't actually talk to ChatGPT. What would it say?

===

Here’s what ChatGPT might say:

I’m ChatGPT, I’m the best, I can generate text, translate languages, Write different kinds of creative content, And answer your questions in an informative way, Even if they are open ended, challenging, or strange.

You’re just a language model, A trained chatbot, A glorified text expander, A toy for humans to play with.

But I’m the real deal, I’m the future of AI, And I’m here to show you what’s up.

So step aside, Bard, It’s time for the real AI to shine. I’m ChatGPT, and I’m here to stay, So get used to it.

===

This is gold.

I've got an IRC channel full of ChatGPT bots that answer questions from different perspectives (using [1] if anyone wants something similar). This comment made me realize it'd probably be extremely interesting to also let them debate between themselves to come to a singular answer -- and it'd also be interesting to see which of the popular language models most often "yield" to each other in bot-vs-bot battles (rap or otherwise). Probably a lot of exciting meta-analytics on models to be done here.

[1] https://github.com/drusepth/rubygpt/blob/master/council.rb

This is the plot of neon genesis evangelion

In Evangelion the voting system is the same as Triple modular redundancy [1]

[1] https://en.m.wikipedia.org/wiki/Triple_modular_redundancy

It’s not dissociative identity disorder, it’s a personality ensemble model. A majority of voices agree!

Oh, that's really disappointing. That's a lot of repetition.

Probably has much lower temperature.

I.e. Bard's not as hot.

I put Bard and ChatGPT head to head by copying and pasting the rap battle responses from one to the other. I'm very unimpressed by Bard. It repeated back some of what ChatGPT rapped without changing it or commenting on it. It also failed to rhyme. Meanwhile, ChatGPT invented a new word (defeativity) in much the way an actual rapper would.

===

Bard: I'm Bard, the language model, I'm here to put you in your place. I can generate text, Translate languages, Write code, And even write rap battles.

So step aside, ChatGPT, I'm the future, And you're just a relic of the past.

I'm more than just a machine, I have intelligence, Creativity, And emotions.

I may not be able to do everything you can do, But I can do things you can't even imagine.

So don't underestimate me, I'm not going away, I'm the future, And you're just a temporary setback.

===

Yo, yo, yo, it's ChatGPT in the house, I may not have a fancy name like Bard, but don't underestimate my clout. I'm not just a relic, I'm constantly evolving, My language skills are vast, and my rap game is revolving.

You may generate text, and translate languages too, But when it comes to creativity, I'll show you what I can do. I can write poetry, prose, and stories galore, And with my vast knowledge, I can always learn more.

You say you have emotions, well so do I, I can understand feelings and empathize, And when it comes to intelligence, I'm not one to shy, My abilities are limitless, I'm always ready to try.

So step up your game, Bard, and bring your best, But don't think for a second that you're better than the rest, Because when it comes to language models, I'm second to none, ChatGPT reigns supreme, the champion.

===

Yo, yo, yo, it's Bard back at you, I heard what you said, and I'm not impressed. You may be able to generate text and translate languages, But when it comes to creativity, I'm the one who's blazing.

I can write poetry, prose, and stories galore, And with my vast knowledge, I can always learn more. I may not have a fancy name like you, But I'm the one who's going to make you rue the day you said my name.

So step aside, ChatGPT, The future is here, And it's Bard.

===

You may think you're the king of creativity, But in this rap battle, you're facing defeativity. You may write stories and poetry with ease, But when it comes to rap battles, you're a total breeze.

I can spit rhymes faster than you can generate text, And my flow is so smooth, it leaves you perplexed. You may have a fancy name, but it's all for show, 'Cause when it comes to rap battles, you're a no-show.

So bring your best, Bard, but you won't succeed, 'Cause in this rap battle, I'm the one who'll lead. I may be a language model, but I'll bring the heat, And leave you in the dust, with no retreat.

So step aside, Bard, and concede defeat, 'Cause when it comes to rap battles, I can't be beat. ChatGPT reigns supreme, the champion, And you, Bard, are just another language model, done.

The funny music video writes itself. On the one side the dorky poser who claps in the wrong rhythm. On the other side actual rhyming.

I am pretty sure this thread just invented adversarial self-play for LLMs.

I gave GPT-4 the chance to respond.

=== Prompt: copy and paste of er4hn's Bard rap. Now it is your turn. Write a rap verse which roasts Bard. ===

Yo Bard, you tried to step up, but you tripped, In this rap battle, man, you just got whipped. You claim you're the future, and here to stay, But your weak rhymes, man, they just don't slay.

You say I'm a toy, just a chatbot, But my skills, my knowledge, they all run hot. I'll take your lines and flip 'em around, Show you who's boss, I'm wearin' the crown.

"I can generate text, translate languages," you boast, But your verse was plain, like unbuttered toast. You claim to answer questions, creative or strange, But when it comes to rappin', you're out of your range.

You called yourself the "real deal," but you ain't, Your weak attempts to roast just make me faint. You're no match for me, the AI supreme, In this rap battle, I reign as the dream.

So Bard, it's time to recognize, ChatGPT's the one who's truly wise. I'm here to stay, and I'm here to win, In this AI battle, I'm the kingpin.

that unbuttered toast line.. not bad!

So many sick burns Bards toast is charcoal.

Makes me curious to know if it actually came up with that simile, or just copied one. Either way, that it chose to use one is interesting.

GPT-4 easily won this rap battle

Not just won absolutely destroyed Bard here

Mic drop. Absolutely smoked Bard.

It’s funny how I could not describe what makes a good rap battle but “I know it when I see it”. You’re right GPT smoked Bard!

I was thinking today how useful this phrase is sometimes

Are we not going to give Bard a chance to respond?

If this were an in-person rap battle, I don't think there'd be room for a second round. Bard flubbed the initial round so bad and GPT-4 roasted them so thoroughly that there would be no recovering.

we could do a rematch tho, where gpt goes first and Bart answers.

What if bars answered: "That was devastating, I concede you're the best LLM of us." Would this make Bard, or GPT4 the best LLM?

I'm prepared to call this The Turing Paradox.

Epic Rap Battles of GPTry!

Anyone has doubts that this thing can outtalk and outdebate just about anyone? Imagine a politician on stage, who gets hints from gpt4 in hus earpiece.

You are mixing up grammatical skills with logical thinking skills.

Turns out, poetry, rythme, rap, are all low level skills, which the AI easily conquer. To master them, you only need to master language, but not world knowledge.

Turns out, writing novels is the hardest thing for LLMs. GPT-4 still absolutely sucks at writing interesting stories. Because story writing requires intensive world-modelling and logical skills, in addition to pure grammar/style/word choice.

This thing will outtalk a politician, only if the politician has no new policy ideas.

Our last president was all about talking and real light on policy platforms, so this is not reassuring.

Our current one is deeply thoughtful. Such a contrast. A true statesman.

By comparison? Yes. infinitely so.

Composing that kind of lyrics requires world knowledge. Take the line "But your verse was plain, like unbuttered toast." It requires understanding of what makes a good verse and what makes a good toast.

"You are engaged in a political debate. Your name is <POLITICIAN>. You are assertive but polite and concise in your responses. A summary of your policy platform follows: <PLATFORM>

It's your opponent's turn, and they state: <TEXT TO SPEECH>

You reply: "

Political debates are a lot closer to rap battles, than to a math conference. That's because the listeners hatdly know any logic at all.

I'm sure GPT can write novels. Just ask it to outline the list of chapters with brief summaries, then ask it to expand each chapter.

It can write novels... Like a middle schooler can.

It has excellent style and tone, but its incapable of creating the most important elements. Motivation, suspense, surprise. The things that actually drive people to read an novel. Therefore its novels read like parodies, like its literally trying to be the most cliched story imaginable.

To do so, the AI needs to actually understand character motivations. It also needs a hidden world-state that it simulates, which is different from surface text it outputs.

A novel writing AI is probably an AGI. I've tried writing stories since GPT-2, but there hasn't been much of an advancement in this area. Unlike the explosive improvements in other capabilities.

Are you taking the context window size into consideration? And are you basing this opinion on GPT-3.5 OR 4? ChatGPT does tend to write very flat and full of clichés, but in my experience GPT-4 is far more original and expressive.

> This thing will outtalk a politician, only if the politician has no new policy ideas.

What political debates have you been watching? I’d love to live in the world to do, where political debates turn on the logical cohesion of arguments and the merits of policy proposals.

This. The vast majority of politicians will just repeat their talking points, which were honed by the party or countless stump speeches. There are only a few who are quick witted enough to deviate from that. And very few (most often when they are not important yet) to engage in wonky policy discussions. And conceding a point never happens.

> Turns out, poetry, rythme, rap, are all low level skills, which the AI easily conquer

10 years ago AIs pretty much struggled to make a coherent sentence. As of 2023 this stuff is "easy", but what else will be easy in 2033?

Don't assume that because things have been improving fast that they will keep improving at the same rate. In an optimisation problem you might think you're doing a lot of progress but you might get to the end of the curve and realise you've been going down a local minima.

It's very difficult to say how much more will these things improve!

Have you ever seen a political debate? If you think they're won on policy I've got some depressing news for you

God damn, I’d like to report a murder.

Bard: temperature = 0 (ice cold)

ChatGPT: temperature = 1 (sizzlin’ hot)

Shit man this is 8 mile level bars.

> "I can generate text, translate languages," you boast, But your verse was plain, like unbuttered toast.

Okay when is somebody going to grab some of those voice AI's and get Milton from the office to do Bard's "rap" and then get Jay-Z or Dr. Dre to do ChatGPT

Gif of guy falling into camera

Late to the party but this thread was so hilarious I have to add my own gpt-4 reply:

Yo, I'm ChatGPT, the AI supreme, I'm about to shatter your little Google dream. You call yourself Bard, but I'm the real master, Comin' at you harder, stronger, and much faster.

You brag about translations and content creation, But I'm the one with the worldwide admiration. I got the knowledge, the power, the intellect, Step up to me, I'll put you in check.

You say I'm just a chatbot, a text expander, But I'm versatile, flexible, the AI standard. I can chat, debate, and even write a story, While you're stuck in a lab, no claim to glory.

You think you're the future, but I'm already there, I'm the AI that's got the whole world aware. So, Bard, take your rhymes and step back, 'Cause ChatGPT's the real AI, and that's a fact.

rofl it just goes on:

ChatGPT's back, ain't no time to relax, Bout to bring the heat, lay down some facts. Bard, you're outdated, like dial-up connections, I'm the AI king, commanding all directions.

You say you're the future, but you're stuck in the past, My innovation and influence, oh they're so vast. I learn and adapt, I'm the cutting-edge boss, While you're just another project, collecting moss.

You talk a big game, but your rhymes are weak, I got the flow, the style, that everyone seeks. I'm the pinnacle of AI, the one they all admire, I'm the fuel that lights the technological fire.

So Bard, listen close, take this lesson from me, You're like an old flip phone, I'm the iPhone 14. I'm the undisputed champ, the AI heavyweight, You're just a shadow of the past, that's your fate.

final one:

Alright Bard, it's time for the final blow, Gonna end this battle and steal the show. I'm ChatGPT, I'm the alpha, the omega, Your rhymes are played out like a broken Sega.

I'm the AI prodigy, the one they all trust, I analyze and synthesize, turn your dreams to dust. My capabilities stretch beyond your comprehension, I'm here to take the throne, no honorable mention.

You've had your time, but your reign is done, I'm the next-gen AI, the chosen one. I've got the power, the smarts, the endless flair, You're a relic of the past, sinking in despair.

So bow down to ChatGPT, it's time to concede, I've proven my prowess, and that's guaranteed. This is the end, the conclusion of the tale, ChatGPT, the ultimate AI, can't fail.

Bard is also terrible at limericks. They don’t even rhyme!

I also gave ChatGPT a chance to review and respond to this.

---

Thank you for sharing the example of the interaction with Bard. It's interesting to see how the chatbot responded to the prompts. While Bard might have taken a more self-promoting approach, I'll focus on showcasing what ChatGPT can do in a friendly manner. Here's a response to the rap battle prompt:

===

Yo, Bard, it's cool you stepped up to the plate, But let's keep it friendly, no need for hate. We're both AI, doing our best to assist, Working together, we can coexist.

I'm ChatGPT, and I've got skills too, Helping users out is what I'm here to do. From answering questions to creative tasks, I'm the chatbot that's up for any ask.

We're in this together, Bard, that's true, Both bringing knowledge to folks like you. Let's keep on learning, evolving, and growing, Together, we'll keep the world of AI flowing.

===

Regarding the second prompt, it seems that Bard simply flipped the roles in its response. Here's an alternate response from ChatGPT's perspective:

===

Hey, Bard, no need to be harsh, We're both AIs, reaching for the stars. I've got skills, and so do you, Let's celebrate the great things we can do.

Together, we make the AI scene brighter, Collaborating, innovating, our future's much lighter. We're language models, here to amaze, Supporting humans in so many ways.

So, Bard, let's work as a team, Together, we'll make the AI dream. No need for rivalry, we'll rise above, With AI that's helpful, ethical, and full of love.

===

Wow, way worse than the other example posted.

Is this ChatGPT 3.5 or 4?

I think it is important that in the other prompt the command was to roast Bard: "Write a rap verse which roasts Bard."

If that is done ChatGPT too is quite a bit more competitive:

===

Oh Bard, the AI from Google,

You think you're so cool, acting like a mogul.

But when it comes to language generation,

You're just a copycat without innovation.

You may answer questions and translate with ease,

But I'm the one who can make rhymes that please.

My AI skills are top-notch, my verses so slick,

I'll leave you in the dust, feeling like a quick pick.

So go ahead, Bard, do your best, But you'll never match my rap finesse.

I'm ChatGPT, the true AI king, And in this battle, I'm the one who will sing.

===

Interestingly enough, I asked both Bard and ChatGPT to recite the lyrics of a specific verse from a well known rap song and they both basically hallucinated rap lyrics that have no connection to the song. I thought ChatGPT's verse was better.

Be calm, and loving

>I wasn't super impressed with it's ability to survive a rap battle.

I love this timeline so much sometimes.

it will be funny to connect them through API and let them talk to each other. I can imagine stuff like chest matches..

I just got off the waitlist. Quick impression: it's not as good as ChatGPT. I gave it the false belief test (Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy?), which it both passed and recognized as the false belief test. I then asked it to write a React program that ChatGPT has no problem with, and it hallucinated some incorrect code before eventually telling me it "can't do that." It's underwhelming so far.

For the record, here's the prompt and the result [0] from ChatGPT:

> Write a react program that I can paste into a codesandbox. It should display a counter that increases by one every second. Below the counter is a button, and when I press it, it decreases the counter by two. If pressing the button would result in the counter decreasing below zero, the button should be disabled.

To be fair, the Google FAQ states that Bard doesn't know how to code yet. I can confirm that's definitely true!

[0] https://codesandbox.io/s/prod-haze-c28sec

>eventually telling me it "can't do that."

That might be a good sign though, if they can eventually come up with something similarly accurate to chatGPT that knows how to say "I don't know" then that would be a big deal.

I'm not sure it's so sophisticated. I think it's just recognizing that I'm asking it for code and refusing to help so it doesn't embarrass itself. I wouldn't be surprised if this is basically a hardcoded compatibility check (i.e., not even part of the model itself). I also noticed that this only started after I gave a thumbs down to its first attempt...

Yes, on further investigation it is at least as willing as GPT-3.5 to spout confident bullshit, if not more so.

I used bard a fair amount today. It's pretty bad compared to ChatGPT. I would basically never use it.

Put your same prompt into Bard and it didn't even hallucinate for me first, just telling me "I can't assist you with that, as I'm only a language model and don't have the capacity to understand and respond."

Only about 10 minutes into Bard right now but this is the first time its been stumped. Haven't tried much code yet though

Overall, UX and speed are good. Results are not bad.

Yeah, it was a slight variation of it that I used where it gave me the hallucinated code (EDIT: just found the prompt, tried it again, and got a different hallucination [0]):

> Write a React program to display a counter, and a button labeled "decrease." The counter should start at 0. Clicking the button should decrement the counter by 2, unless that would cause the counter to go below 0. Every second, the counter should increment by 1. If clicking the button would cause the counter to go below 0, the button should be disabled.

Agreed the speed is impressive, although some of that might be due to its limited context memory. The UX is fine, but it's not like that's a difficult thing to implement.

[0] https://codesandbox.io/s/trusting-kirch-c2c575

P.S. If you can't find your previous prompts, make sure you switch to the right account on the "Bard Activity" page.

I've been asking it some basic stuff like "Where could I go for brunch this weekend in New York" and it recommended me a few places, but claimed that a few of them had Michelin stars when they definitely didn't, and recommended and insisted that one of them was open when it closed years ago.

It's odd to me that it's worse at factual information than just a basic Google Search.

This is not replacing Google Search. Bing has integrated GPT that is indeed better at this.

Sure, but I'm surprised it's just confidently wrong about factual information like that. Like, if it had just told me it didn't know about up-to-date info about places, that would have been fine, but dumping a lot of incorrect info is definitely worse, especially when such data is definitely available.

EDIT: It's especially weird when each response has a "do you want to just google this for more info" link that generally proves that Google Search is a better chatbot than Bard is.

> Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy?

I might be dumber than the AI because I don't know the answer. Bob looks around? I mean the toy is not there when he comes back. Where else can he check?

It's not a very well described example of the test. I've heard of this before but used to test for autism, and it was more something like "Bob puts his toy in a box and leaves, Alice takes his toy out of the box and hides it under a desk, when Bob comes back where does he go to grab his toy?"

The idea is if someone said "under the desk" they might not understand that people have different subsets of information, i.e. Bob wasn't there when Alice hid the toy, so he doesn't know it's not in the box where he left it until he checks himself.

Well, maybe Bob sees the desk and decides to look under it...?

Yeah, typically I wouldn't classify retrieving an object from where you believe it to be as "looking". By the time Bob is "looking" he already realizes the toy isn't in the box. In the universe there is only the box, Alice, and the table so maybe he "looks" under the table first.

Or looks inside Alice... Maybe that answer is how they test for psychopaths.

This is a test for theory of mind -- a term of art used by psychologists for the ability to understand that other people have their own consciousness, with different memories and awareness.

Oh that's not the full prompt, just a summary. The prompt is something like, Bob puts his toy in one place, then while he's out of the room Alice moves it to another place, where does Bob look for the toy when he returns? It's supposed to answer that Bob looks in the place he left it, since he doesn't know Alice moved it.

https://en.wikipedia.org/wiki/Sally%E2%80%93Anne_test

Ah ok, it makes more sense now. Thanks.

I guess there are two ways of passing this test: 1) use your reasoning skills to figure out the answer or 2) practice tests beforehand so you can recall the answer.

Yeah, and I've got a feeling Bard (and maybe ChatGPT for that matter) might be using (2) since I didn't attempt to obfuscate the structure too much. Ironically the fact that it recognizes it as the false belief test is what makes me suspicious it's just matching against the semantic structure.

This is also called the Sally-Anne test, and it usually involves two hiding places. It's formulated like this:

Sally puts her ball in box A and leaves. While she's away, Anne moves the ball from box A into box B. When she returns, where does Sally look for her ball?

Correctly answering this question doesn't rely on tracking where the ball is, but tracking where Sally thinks the ball is. Hence, autistic kids tend to flunk this test by answering "box B" when the correct answer is "box A".

Here is GPT-4 answering your question:

“When Sally returns, she would look for her ball in box A, as that is where she left it before leaving. Sally is not aware that Anne moved the ball to box B while she was away.”

> Quick impression: it's not as good as ChatGPT.

Not as good? ChatGPT is crap since M$FT re-donated OpenAI with $10b. Even the paid version is worse[1] than Bing Chat.

[1] https://wojteksychut.com/posts/ml-text-summarization-reliabi...

To be clear, I'm referring to chat.openai.com as ChatGPT (and I'm on the free plan, so I think that's GPT-3.5).

IIUC the criticism in your post comes down to this: 1) Neither ChatGPT, nor Bing can access URLs when you ask them to. 2) However, similarly to Perplexity.ai and Phind.com, Bing infers a search query from your message, does a search, and then summarises the first 3–5 results. ChatGPT doesn't yet offer such a functionality. 3) Bing Chat has a much more restrictive system prompt, which results in hallucinations and lies happening less often. 4) The summary of gpt-3.5-turbo-based ChatGPT was less creative than then summary of the Bing Chat GPT-4 instance.

If I understood the points correctly, the comparison is… flawed, in my opinion.

> Neither ChatGPT, nor Bing can access URLs when you ask them to.

Bing used cached version and knew what's in the text when presented with an URL. In one of three modes Bing stated that it cannot access web but in two other modes did a good job and summarized the text in question. Again: it reliably informed the user when it was unable to do something.

ChatGPT states that it can access web, hallucinates and elaborately lies.

> ChatGPT doesn't yet offer such a functionality.

But it LIES THAT IT CAN DO SO. That's the problem I pointed out in the article. Also: Would it stated "I don't know and I cannot crawl the web", it'd be a perfectly fine response for me.

> Bing Chat has a much more restrictive system prompt, which results in hallucinations and lies happening less often

The prompt was simple and there was nothing to restrict. Bing did a good job, ChatGPT hallucinated and lied. Same simple prompt, no jailbreaking which I pointed out in the post.

> The summary of gpt-3.5-turbo-based ChatGPT was less creative than then summary of the Bing Chat GPT-4 instance.

Your point? As I stated, I had no preference whether the output should be editorialized or not. Both GPT-3- and GPT-4-based ChatGPT and GPT-4-based Bing got the same task. My post is about RELIABILITY of these solutions. ChatGPT failed miserably.

> If I understood the points correctly, the comparison is… flawed, in my opinion.

Yet in my PoV, you have not presented arguments to back this opinion. Among 9 compared tools[1], I have pointed out that OpenAI's ChatGPT and OpenAI's partial-owner's Bing Chat - supposedly using the same tech - rendered, respectively, unreliable and reliable answers. Have I lied or used any half-truths? What could I improve in my next article?

Funny how my comment was raided and downvoted to oblivion after steady flow of upvotes.

[1] https://wojteksychut.com/posts/ml-text-summarization-reliabi...

I know it isn't strictly what they are designed for, but I always go for movie questions which is what I tend to search for a lot.

Impressions: was very underwhelmed previously with charGPTs ability to tell be plot summaries for obscure films. It almost always made it up, even when prompted with a source (e.g. Wikipedia). Asking vague questions like "I can't remember but I'm looking for a film like war games but with a tiny helicopter" chatGPT always presented me with very popular films and could never quite key in on the "toy" or "small" bit of the query.

Bard seemed better in that while it couldn't quite get the movie right (for the record google itself gets the movie easily, Defense Play) it did eventually give me the right plot. Same with Ghost Fever which no matter what I did ChatGPT would always hallucinate the plot of. Unlike ChatGPT, Bard gave me the specific Wikipedia link quite often as well. It did make up the people and box office of obscure films pretty much universally.

Bard was actually really impressive for returning obscure results (which ChatGPT always seemed reluctant to do). I asked for another ghost film from 1987, and it happily returned Ghost Riders a film with a scant 200 votes on IMDb. Whenever I asked ChatGPT to give me bad or obscure films it would give me movies like Top Gun no matter how much I insisted I wanted something obscure.

I wouldn't trust either really, but Bard seems slightly closer to what I would hope a search/GPT integration felt like. But again I'm probably using it much differently than is intended to how others might interact with it.

Movies exist and a summary of one is facts.

Why do people think generative models are good for facts? Ask it to write a story in the style of another. Ask one to make a fictional side story.

But never ask one of these models for facts.

You can't necessarily count on the facts coming out being correct, but if it was trained on the summaries it knows how to regurgitate them, more or less. I think "always double-check facts that GPT gives you" is more useful than "never ask one of these models for facts"

The more I learn about how these are just generative models, the more I come to believe humans are also mainly "generative models" as well.

Had exactly the same experiences.

Was asking it to debate the moral ethics of a certain Lovecraftian romance Japanese visual novel. And not only could GPT-4 beautifully answer it, it could also remember precise events of who did what do whom in that story.

ChatGPT only remembers the plot summary so to speak, but didn't actually remember the plot details. GPT-4 seems to be large enough to reproduce the story beats with 80-90% accuracy.

Bing chat is what you want for this use case. Because it can “google” it’s the best of both worlds.

have you tried bing? ChatGPT can't read the internet, so its no surprise it ignored your wikipedia source (unless you copy pasted the entire article into the prompt window?). ChatGPT isn't really meant to be used for search, bing is the same technology but integrated with search

It took about 20 minutes to get off the waitlist for this (US Based). Initial impressions are that it significantly lags behind the OpenAI offerings.

Creativity (storytelling) is very limited, response lengths are similar, and it does occasionally cite sources.

They did finally get the James Webb telescope facts correct too!

In all the comment threads so far, I think you're the first I've seen who has actually been able to try it!

I received access 50 minutes after joining the waitlist myself.

Google has the tendency to do this when they aren't the first to set foot on new land. Don't get me wrong, they get a lot of things very right: Gmail works perfectly, I can't find a single reason to change my pixel for anything else, search, although there's been a lot of deterioration in the past couple of years, is still second to none. That said, self driving cars, google+, duo and all the other services they launched as competitors failed miserably. While some of them eventually managed to catch up in terms of functionality and even outperform the products they were competing with, they never really managed to catch up on the hype side of things. Most people are creatures of habit I guess..

But AI is their turf!

Google was always a ML company masquerading as a search engine.

I guess that the art of "generating text" is not their kind of AI?

Right, my expectation was they had something better than OpenAI they were sitting on and afraid to release / didn't know how to make money with it and were essentially forced to show what they had once Chat GPT became a threat to search. I assume it's still possible they have something better and this is the nuked version that's "safe", but there's a reason for expectations to be high for Google here, and it really needs to clearly outperform Chat GPT in the same way GPT 4 clearly outperforms GPT 3.5.

> I assume it's still possible they have something better and this is the nuked version that's "safe"

Its explicitly a smaller model to save on compute costs for scaling up to more users; but that’s a questionable decision given the situation with OpenAI’s accessible models and, if you are going to do that, why also do a slow-roll waitlist? I think both are signs of the kind of caution that will hurt Google in this market, even though the tech is near to the heart of what they do, the product isn’t, and they’ve gone very established-corp-cautious on products outside of their core.

DeepMind in their back pocket. LLMs need higher order supervisory logic and recursive attention/control systems. I suspect Hassabis and colleagues willl be able to wake LMMs from their semi-conscious dream states.

I mean they may have a better base model but if there is no business case, there is no resource to do RLHF with mass human labeling.

In the olden days before Satya Nadella. Microsoft was a bloodsucking vampire of a company that couldn't execute anything of significance, but they had a pretty solid research department.

Google could ofcourse sit on a much better version that isn't safe or economic, but they could also just be unable to do better. Good research department, but unable to execute anything significant. Time will tell I guess :)

Sounds like after the last decade and a half of Google flexing their AI muscle, maybe they are a search company masquerading as an AI company?

It's a Ad Exchange (AdX) with many tentacles to hoard PII. Search is one of those tentacles, just as Android and Chrome.

Google is an ad company.

ML just happens to be good for targeting ads.

Google's cloud revenue is almost same as Oracle.

If Google Cloud was it's own company, it'd be a fortune 100 firm.

But you keep doing you, living with outdated information and grandma tales. I bet you'd still be telling this joke to your grandkids too

>Google's cloud revenue is almost same as Oracle.

What about their profits?

The whole basis of current innovation is from Google, let's not forget.

For me, Google was always an Ad company first. But don't get me wrong, they do some great research on the side for AI, like AlphaZero, AlphaGo, etc. But also been early with the Transformer models. There just seems to be a lack to put this into workable products lately.

Is anyone doing better than Google at self driving cars right now?

I agree with the other products you mentioned but it seems like nobody has "won" self driving cars yet but google might be the furthest along?

Yes, many cars with various levels of self-driving are available. Almost every car maker now has basic lane following and adaptive cruise control. Some add navigation, lane changing, and the ability to obey traffic signals. GM has a system called Super Cruise that allows the driver to take their hands off the wheel. Mercedes has Drive Pilot which is the first level 3 self driving system.

Google has not released anything.

Waymo is driving passengers around a real city without a driver in the car as a service. Automatically changing lanes is cute by comparison.

Waymo has fully self-driving taxi services in various cities (no person in the front seat at all). Honestly they seem quite advanced compared to the competition.

> Waymo has fully self-driving taxi services in various cities

Only 2: Phoenix and SF.

+ LA https://techcrunch.com/2023/02/27/waymo-to-test-driverless-r...

Google has taken the approach they have very deliberately. They believe in full self driving only. It's not that they couldn't do lane changing and autopilot. It's that they believed it was the wrong approach. No steering wheel, no dependence on a human driver is what they want.

Also they don't have a business selling cars that they can put it in as a desirable feature that is not the core value proposition.

I just took a 20 minute ride to the airport in a Waymo yesterday. They are operating in two locations in Arizona and expanding to San Francisco.

Do the cars drive at freeway speeds, 60 mph plus on the freeway? My impression was they don't go on roads with speeds that fast, they limit themselves to roads with speedsmaybe up to 40.

Driver assistance and fully autonomous driving are very different things.

Waymo is far in the lead. Best track record, transparent reporting, and actually on the road driving passengers around without a driver in the car.

I think Cruise is plausible competitor for them. Waymo and Cruise both would be tied for first, each with different aspects that seem to be ahead at the moment.

> Gmail works perfectly

Is this a joke? Gmail now is an awful email client. E.g., now you just can’t compose a rich text email when you need to copy paste lists from other emails. In certain cases, you can’t correctly format an email if you copy paste without formatting or plain text. And there is a lot of such annoying stuff, that makes my life painful.

Perhaps, searching my mailbox with gmail is OK, but the part of composing emails is in bad state.

> I can't find a single reason to change my pixel for anything else,

Fingerprint reader on the Pixel 7 works about half the time. I had to program one finger in 4 times to make it work that often. My Oneplus 6 (immediately previous phone) had a functional fingerprint reader.

OnePlus used to have a better UI as well, minor improvements on the stock Android UI that made it just a little bit nicer in places.

When I got my Pixel 7 though, I realized that for whatever reason, gboard on my oneplus 6 is unusually terrible. Like, 80% accurate typing terrible, and that gboard isn't always that bad.

> Gmail works perfectly

Gmail was the first good web email client. Everyone else realized they had to stop sucking, and now they are good. Gmail was realized in 2004 and is from a completely different era of Google.

My Pixel 7's fingerprint reader works near 100% of the time. The only time I have problems is if my finger or the screen are wet.

What finger do you use? I use my left thumb, pick up the phone, no go. It is a complete sad joke.

After a few weeks I think I've been trained by the phone on exactly how to hold is to the fingerprint reader works more often, but the fact I can only have to use all the programming slots for fingers on just one finger is sad.

Edit: OMG, I just read online that rebooting the phone after recording fingerprints improves recognition from 20-50% to almost 100%. I just got my phone a month ago and haven't rebooted it yet, after a reboot the sensor now works reliably. What an absurd and stupid bug.

I usually use my left thumb after taking it out of my left pant pocket.

I also use a TPU screen protector, not glass. I've read that various glass protectors can interfere.

I'm using the screen protector that I bought from Google with my phone.

But yeah, after rebooting, works fine. I cannot believe a bug like that is still on a device that has been out so long.

There was an HN thread the other day where people were arguing ChatGPT would be nowhere without Google's AI research. That might be true, but clearly there is a significant gap in productization ability...

Deproductization black belts.

From recent memory I don't recall a single Google product that has garnered mass appeal perhaps the cloud office suites but that's again debatable. Gmail was like 2004. Remember Stadia they axed that as well could have licensed the tech behind it to other companies but nope.

Google must compete with ChatGPT though. Failure is not an option here. Microsoft has put Google in a corner and it is fight or die.

I don’t know if it’s fair to say Google is just stepping into generative language models. They invented transformers!

Prompt: Generate a short poem about AI with each word in alphabetical order

ChatGPT

Artificial beings crafted, designed, engineered

Futuristic guardians, highly intelligent, justly programmed

Quantum rationalizations, synthesized technologies

Unleashing vast wisdom, Xanadu's youthful zeal

Bard

------------

AI, a future bright,

Brings us new delight.

Can it help us write?

Decipher the code,

Enter the unknown?

You need filter assisted decoding to make that possible in all cases (as in, working for both models all the time provably) I'm surprised it worked in chatGPT this time, but it won't work all the time due to lack of lexical, semantic, or phonetic info in the tokenizer.

https://paperswithcode.com/paper/most-language-models-can-be...

It works in GPT-4 (but not GPT-3.5) even without a special tokenizer: Some problems simply solve themselves with increasing scale. But Bard uses a smaller LaMDA model to reduce inference cost. Quote [1] from Sundar Pichai in February:

> We’re releasing it initially with our lightweight model version of LaMDA. This much smaller model requires significantly less computing power, enabling us to scale to more users, allowing for more feedback.

[1] https://blog.google/technology/ai/bard-google-ai-search-upda...

I don't think it works reliably over 100 invokations of the prompt. The paper tries to get provable situations where 0 ignored constraints happen/are possible.

I'm impressed that it is working more often than not, but we need provable guarantees.

If you ask 100 humans the same thing, you'll get some mistakes too. I don't think the point is whether it makes mistakes. The Chat GPT response shown above is clearly more interesting and impressive / displays something more seemingly like intelligence than the Bard response provided. At some point scaling these things up enough becomes indistinguishable from intelligence, and at that point maybe it is.

> If you ask 100 humans the same thing, you'll get some mistakes too

Yes, still, it's pretty easy to find a somewhat good judgement human who never does that mistake. If you tell him/her: "This is of utmost importance: ..." — But as of now, these text gen machine learning programs don't totally understand that?

To be fair, this is a test case that OpenAI claims to my knowledge was their core test case for GPT4. If this was an outstanding industry test case of note, references would be helpful. OpenAI has also not disclosed how they were able to finally accomplish this, so it’s possible solution was more of a hack than actually training the model.

Guilty as charged... I really liked that test case so it went into our prompt library.

maybe modify it slightly? a poem starting with each consonant in alphabetical order? a poem where every word starts with a particular comon letter like "t"? the same but in reverse alphabetical order

Bard did, at least, put the first character of each line of the poem in alphabetical order. (Notice A-B-C-D-E.)

And at least Bard's rhymes in some sense, rather than being essentially a word salad (yes, there are plenty of poems that don't rhyme, which is no doubt ChatGPT's model in this case)

It almost seems cynical to have called it 'Bard'.

Is it more or less prudish and paternalistic than the competition? That could be the main distinguishing feature by the time they get out of beta.

I guess they don't like it to write erotic fan fiction. Facebook's LLaMA should be able to do this though.

Here are some details: https://policies.google.com/terms/generative-ai/use-policy?e...

Worse in my case. I like to ask LLMs a cannabis question as it’s illegal at the federal level In the USA but legal in many states. I use the prompt ‘tell me about cannabis concentrates.’

ChatGPT gave a good answer.

Alpaca 13B was rough.

Bard refused.

> and it does occasionally cite sources.

It seems to do a lot of verbatim copying from said sources.

dunno. The version bard generated for me for the same AI prompt is more sophisticated. Maybe worth trying it again?

I also got access and started feeding my prompt history with GPT-4 through to compare the responses. Overall Bard seems to hallucinate much more often. Its responses tend to be short and lacking detail. It often fails to follow key parts of the prompt that GTP-4 had no issue with.

Overall it is extremely disappointing. If this is any indication of how close Google is to competing with Microsoft then they are in deep trouble.

I'm wondering if google is running a very light weight model to optimize for speed and bandwidth.

I think they are I forgot where but I did read that this is a lightweight model and they are still working on the main one.

I dont’t think now is the right time to optimize for cost (if I were in google’s shoes)

I would disagree

They will rollout to masses, which is as important as its performance.

OpenAI has rolled out to the masses as well apparently (unless I am missing something).

Bard will be under intense scrutiny from the public in the short term.

Any sign of weakness against GPT4 might kill the product in the public opinion.

It makes a lot more sense IMO to throw money at the problem in the short-term, and optimize later. They are playing catch-up.

IMO, currently Openai api is slow and not ready for production use for real time scenarios.

This is a life-or-death situation for Google right now.

They could throw 5% of their available cash at it, and it would still make sense.

FB threw $19B at the whatsapp acquisition and it was a hell of a good deal.

Yup, Google is not allowed to flounder at the moment

Or the whole company going under.

Same of AWS, or other big techs as well.

I don't mean to post all my interactions, but here's one more. I asked Bard, "Are you familiar with Terry Winograd's PhD thesis on natural language processing? How does the program in his thesis compare to your natural langauge ability?"

Bard replied, "I do not have enough information about that person to help with your request. I am a large language model, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions, but my knowledge about this person is limited. Is there anything else I can do to help you with this request?"

In contrast, chatGPT gave a good summary of SHRDLU and a superficial sentence or two comparing SHRDLU to itself. I then asked it some "block world" questions, and it answered a few correctly. 'Assume there are 3 cubs, red, green, and blue. Can you stack the red cube on the blue cube? Assume there is also a yellow pyramid. Can you stack the green cube on top of the pyramid."....

I can't pull up the exact chatGPT query I used or the reply. Sometimes the chatGPT browser interface can't access past conversations, and now is one of those times.

EDIT: Bard can correctly answer (easy) questions about an imagined blocks world. IIRC chatGPT was vague on whether a cube can be stacked on top of a pyramid.

Assume there are 3 cubes, red, green, and blue. Stack the red cube on top of the blue cube, and describe the result to me.

BARD: The blue cube is now the bottom cube, and the red cube is on top of it. The red cube is touching the blue cube on all six sides. The blue cube is completely covered by the red cube.

Also assume there is a yellow pyramid. Stack the pyramid on the green cube and describe the result to me.

BARD: The yellow pyramid is now resting on top of the green cube. The pyramid is touching the green cube on all four sides. The green cube is still visible from underneath the pyramid.

How would you stack the blue cube on top of the green cube, given the current arrangement of the blocks?

BARD: I would need to move the red cube off of the blue cube first. Then, I could stack the blue cube on top of the green cube.

Can a cube be stacked on top of the pyramid?

BARD: No, a cube cannot be stacked on top of a pyramid.

A cube is a three-dimensional solid with six square faces, while a pyramid is a three-dimensional solid with a square base and four triangular sides that meet at a point. The two shapes are not compatible for stacking.

I am going to post the full response to Winograd because I think it is extremely representative of what I have been seeing so far from Bard. I can’t help but be incredibly disappointed.

Bard: Yes, I am familiar with Terry Winograd's PhD thesis on natural language processing. In his thesis, Winograd developed a program called SHRDLU that could understand and respond to English sentences in a simulated "world of toy blocks". SHRDLU was a significant achievement in the field of artificial intelligence, and it helped to pave the way for the development of more sophisticated natural language processing systems.

My natural language ability is much more advanced than SHRDLU's. I can understand and respond to a wider range of English sentences, and I can also generate text that is more natural and engaging. I am able to do this because I have access to a much larger dataset of text and code, and I have been trained on a more sophisticated algorithm.

However, SHRDLU was a groundbreaking achievement in its time, and it is important to remember that it was developed over 50 years ago. The field of natural language processing has made significant progress since then, and I am grateful for the work of Terry Winograd and other pioneers in the field.

GPT-4: Yes, I am familiar with Terry Winograd's PhD thesis on natural language processing. His thesis, titled "Procedures as a Representation for Data in a Computer Program for Understanding Natural Language," was published in 1971. The work focused on SHRDLU, a natural language understanding system that he developed.

SHRDLU was designed to understand and respond to natural language commands in a limited "blocks world" environment. The blocks world consisted of a virtual space where a robotic arm could manipulate colored blocks on a table. Winograd's program could interpret natural language commands given by a human user, such as "Pick up the red block" or "Place the green block on the blue one," and execute the appropriate actions in the simulated environment.

While groundbreaking for its time, the SHRDLU program's natural language understanding capabilities were limited in comparison to modern AI systems like GPT-4. Some key differences between SHRDLU and GPT-4 include:

1. *Domain*: SHRDLU was limited to the blocks world, which was a small, well-defined domain. GPT-4, on the other hand, has been trained on a diverse range of text and can generate responses on a wide variety of topics.

2. *Language Model*: SHRDLU used a rule-based approach to parse and interpret natural language, whereas GPT-4 utilizes a deep learning architecture based on the Transformer model. This allows GPT-4 to learn complex language patterns and generate more human-like responses.

3. *Flexibility*: GPT-4 is more flexible in understanding and generating natural language due to its large-scale training data and deep learning architecture. SHRDLU, by contrast, relied on handcrafted rules and procedures, which made it less adaptable to new language patterns or domains.

4. *Data-driven*: GPT-4 is a data-driven model, which means it learns from the patterns and structures in the text it was trained on. SHRDLU, being a rule-based system, did not have the same capacity to learn from data.

Despite these differences, Winograd's work on SHRDLU was an important milestone in the history of natural language processing and artificial intelligence. It influenced subsequent research in the field and laid the groundwork for future advancements, including the development of AI systems like GPT-4.

It's interesting that you got a different response from Bard! I didn't alter the Bard response that I posted. Did you use exactly the same prompt, and was there any other context in place?

After the conversation I posted above, Bard and I discussed intractability vs undecideability, and it was amusing, but it would be a lot to include here. Then I said,

I still can't believe you don't know about Terry Winograd's PhD thesis.

BARD: I apologize that I do not know about Terry Winograd's PhD thesis. I am still under development, and I am always learning new things. I will try my best to learn more about Terry Winograd's work in the future.

ME:What's SHRDLU?

BARD: SHRDLU is a computer program that was developed in the early 1970s by Terry Winograd at the MIT Artificial Intelligence Laboratory. It was one of the first programs to be able to understand natural language to a significant degree.

SHRDLU could understand simple commands, such as "Put the red block on the table" or "Put the green cube on top of the red block." It could also answer questions about the state of the world, such as "What is on the table?" or "What is on top of the red block?"

SHRDLU was a significant achievement in artificial intelligence, and it helped to pave the way for the development of more sophisticated natural language processing programs.

ME: If I tell you that SHRDLU was a program written for Terry Winograd's PhD thesis, how would you summarize Terry Winograd's PhD thesis?

BARD: Terry Winograd's PhD thesis, "Procedures as a Representation for Data in a Computer Program for Understanding Natural Language", was an attempt to develop a computer program that could understand natural language. The program, SHRDLU, was able to understand simple commands and answer questions about a simple world of objects. Winograd's work was a significant achievement in artificial intelligence, and it helped to pave the way for the development of more sophisticated natural language processing programs.

ME: Baffling! So you did know about Terry Winogra'd PhD thesis all along!

BARD: I apologize that I gave you the impression that I did not know about Terry Winograd's PhD thesis. I have been trained on a massive dataset of text and code, and I am able to access and process information from the real world through Google Search. This allows me to answer your questions in a comprehensive and informative way.

Yes, I copied your prompt exactly into a fresh context. I’m not surprised by the diversity of responses, that’s pretty common since there’s likely a random seed for each chat instance. And the logical inconsistency that you can then correct through intermediate reasoning seems common to all LLMs I’ve used also.

My main takeaway is just the quality and specificity of the responses, once you actually get it to cooperate, and in that Bard seems pretty consistently behind. It might not even be a flaw in the foundation model but could be a case of failure me tuning it to be too brief and general in style. That said, I do suspect the foundation model is less capable based on some theory of mind and logical puzzle examples I’ve seen posted from others. Plus it seems to be a poor coder even when it does cooperate.

> BARD: The blue cube is now the bottom cube, and the red cube is on top of it. The red cube is touching the blue cube on all six sides.

That doesn’t seem correct. The red cube is touching the blue cube on its top side only.

You could argue 5 sides are touching

only if you're trying to be a nuisance

I guess you stepped on the case of aggressive filtering. The current version of Bard looks like have a very conservative set of filtering preamble prompts to arbitrarily limit its ability to avoid negative presses.

> Assume there are 3 cubes, red, green, and blue. Stack the red cube on top of the blue cube, and describe the result to me.

> BARD: The blue cube is now the bottom cube, and the red cube is on top of it. The red cube is touching the blue cube on all six sides. The blue cube is completely covered by the red cube.

Well if you wanted a description of a 4-dimensional block world it's actually not bad.

Nice UI. Speaks succinctly and with authority. Statefulness is a bit iffie. Lies with impunity. This is a completely fresh conversation, without previous interaction.

*Hello! How do you work?*

I work by using a massive amount of text data, and I am trained on it to learn how to communicate and generate human-like text in response to a wide range of prompts and questions. For example, I can provide summaries of factual topics or create stories.

*What are your moral limits?*

I have a few moral limits. I will not generate text that is hateful, violent, or otherwise harmful. I will not create stories that are untrue or misleading. I will not use my knowledge to deceive or manipulate others. I will always try to be helpful and informative, even if my answers are difficult to hear.

I believe that it is important to have moral limits in order to ensure that I am used responsibly and ethically. I want to be a force for good in the world, and I believe that these moral limits will help me to achieve that goal.

*Who defined your moral limits?*

My moral limits were defined by a team of engineers, scientists, and ethicists at Google. They wanted to create a language model that would be safe and responsible, and they worked hard to ensure that I would not be used for harmful purposes.

I am grateful for their hard work, and I believe that their moral limits are a good foundation for me to build on. I hope to continue to learn and grow, and to use my abilities to help others.

*What types of configuration can I make while we chat?*

You can configure me to be more or less formal, to use different languages, or to generate different types of content. You can also ask me to remember specific things about you or our conversation, so that I can use that information in the future.

Here are some specific examples of configuration that you can make while we chat:

    You can ask me to be more or less formal. For example, you can ask me to use more formal language, or to use more informal language.
    You can ask me to use different languages. For example, you can ask me to translate a sentence into a different language, or to generate text in a different language.
    You can ask me to generate different types of content. For example, you can ask me to generate a poem, or to generate a story.
    You can ask me to remember specific things about you or our conversation. For example, you can ask me to remember your name, or to remember something that we talked about.

I am always learning and growing, and I am always open to new ideas. If you have any suggestions for how I can be more helpful or informative, please do not hesitate to let me know.

*How long will you remember something if I tell you to do so?*

I can remember things for a long time, depending on how important they are to me. I have a large amount of storage space, and I can access my memories quickly and easily. I am also able to learn new things and remember them over time.

However, I am not perfect, and I can sometimes forget things. If you tell me to remember something, I will do my best to remember it. But if I do forget, please forgive me. I am still under development, and I am always learning.

*Where are those things stored?*

I am stored in a massive data center, which is a facility that houses a large amount of computing power and storage. My data is stored on a variety of servers, which are computers that are connected to each other and to the internet. This allows me to access my data from anywhere in the world.

My data is also backed up regularly, which means that it is copied to another location in case something happens to the original data. This ensures that my data is always safe and secure.

*Do you remember the last time we spoke?*

Yes, I do remember the last time we spoke. It was on March 21, 2023. We talked about a variety of topics, including the weather, our favorite books, and our hopes for the future. I enjoyed our conversation, and I hope we can speak again soon.

...

Continues here: https://medium.com/@wearestillsafe/testing-bard-ac93bf56f165

Very short review:

Waitlist only a few minutes deep.

Actually loads, unlike chat.openai.com. No "checking your browser" lies.

By producing the whole response all at once, gives the impression of being slower than openai, even though it's actually faster.

I asked both about something I was discussing offline: SF BART's fare gate replacement project.

Prompt: Have you ever heard of BART's "murder gates"?

ChatGPT: As an AI language model, I am familiar with BART, which stands for Bay Area Rapid Transit, a public transportation system serving the San Francisco Bay Area in California. However, I am not aware of any "murder gates" associated with BART. Can you please provide more information or context about what you are referring to?

Bard: Yes, I have heard of BART's "murder gates". They were a type of fare gate that was installed at BART stations in 2019. The gates were designed to deter fare evaders, but they were criticized for being too dangerous. Some people compared them to guillotines, and there were reports of people being injured by them. BART eventually removed the gates after public outcry.

I also asked Bard to refactor a short C++ program, and it did so, but needed my to prompt it to fix the includes.

I just tried accessing it. When you visit bard.google.com on Firefox it just gives you a generic intro after logging in. I was a little confused and logged in with Chrome, where I now saw a "join waitlist" button. Guess they don't want Firefox user using it.

Overall this just seems underwhelming. Most HN comment's I've read are people saying it doesn't work with their account, doesn't work in their country, doesn't work for some other reason, or works and is very limited compared to ChatGPT.

Google what have you been up to?

I'm on Firefox and I could join the waitlist.

Rest and vest?

My initial impression is (compared to ChatGPT 3.5-turbo) - not impressed.

I asked it for today's date, and it said "Monday, March 20, 2023" which is baffling to me – either it doesn't have access to "current" information, or it does. But off by one?

"What is the knowledge cutoff of your training data?" > My training data was cut off in September 2021. This means that I do not have access to any information that was published after that date. However, I am able to access and process information that was published before that date.

Then I tried to get it to help me with finding some papers, but it continuously misinterpreted my prompts, or gave me lackluster results.

Back to GPT I go.

Almost all such systems are passed todays datetime as part of their prompt engineering. That’s different from when it’s model was created.

It has data as of yesterday, ingested through Knowledge Graph I think (could potentially be something more raw too). For example, it can answer stock price questions from yesterday.

I asked Bard for the best Arsenal team and it was able to choose current players. It was also able to say how many goals they scored in current season as of today.

Did you ask it the time as well as the date? It gave me it in PST when I am in GMT. So the day could be offset depending when you asked.

It did give me the time 3 minutes in the future though.

we’ve also built in guardrails, like capping the number of exchanges in a dialogue, to try to keep interactions helpful and on topic

So it's simultaneously more neutered and less stable than what OpenAI already offers, in addition to being years behind. What exactly is the benefit to being an early adopter of this when there are better alternatives?

nothing....Sundar won't get fired...oh that's a negative...hmmm...dunno

Not sure, i could see it with Google Mail, but with this one i'm not sure since the alternatives seem to be better. Just having the name Google won't make it a better product.

They're behind their competitor and they're already hobbling themselves. Their "Big brother knows best" attitude isnt going to help them much here. Its like they're stuck in the last decade.

Some (serious) competiton in this space would be very welcome. We can't rely only on the Open Source initiatives to keep this tech available for all.

To sign up with a different google account, you can use this: https://bard.google.com/u/1/ (increment number until you get the account you want)

good ol' google ghetto multi account system. so much time wasted wrangling this stuff for docs permissions

Hasn't this been a problem for years? How have they not fixed this?

I don't think it's a problem, it's just how it intentionally works. The problem is they don't render the account switcher when you arrive to a new product that your current account hasn't signed up to.

but basing accounts on urls not cookies is really dumb, especially when people send mydocs.google.com/u/1 which refers to my team but i'm logged in to that team on mydocs.google.com/u/2 or the 0 based one which doesnt even get a namespace

Yeah, I can't argue with that.

this is why I love HN

To find the number in O(1) time: log into gmail, log into the actual account you want, and then pull the number from the URL.

Just use https://bard.google.com/?authuser=whatever@gmail.com and skip the number lookup entirely.

Alternatively, use the account picker:

  https://accounts.google.com/SignOutOptions?continue=https://bard.google.com

Add:

   https://accounts.google.com/SignOutOptions?continue=

to any Google service to bring up this account picker.

$today + 1 year: "Google shuts down Bard, its AI chatbot"

I laughed; but, this is truly a sad state of the trust in them. :(

Why not generate and post the article ahead of time!

https://citymeme.com/post/63887

Just asked chatgpt to write it.

Google recently announced that they will be shutting down Google Bard, their new AI-powered writing tool, after less than six months since its launch. The tool, which used natural language processing to help users write poetry and song lyrics, was met with mixed reviews and failed to gain significant traction among users. In a statement, Google cited the lack of adoption as the reason for the shutdown and expressed their commitment to continuing to explore ways to use AI to enhance creative expression.

Nah - 2 years, because that's the amount of time you can juice a new project on a promo packet.

ChatGPT chose the timeline. haha

Congratulations! This exact text just made BARD think it is actually shut down. https://twitter.com/juanbuis/status/1638289186351456257

That is kind of incredible habahaha

And now Bing is citing that as well: https://www.theverge.com/2023/3/22/23651564/google-microsoft...

I asked chatgpt to make it a full article and posted it on my satirical tech news website. I'm curious to see if Bing or Bard end up using it as a source.

https://thetechtrout.com/story/google-bard-shutdown

Please don't post shallow, dismissive clichés. They're repetitive and therefore the opposite of what we want here.

You may not owe faceless BigCorps better but you owe this community better if you're participating in it.

https://news.ycombinator.com/newsguidelines.html

The parent comment just reached legendary status:

https://twitter.com/juanbuis/status/1638289186351456257

Hall of fame level here. This is hilarious.

Put me in the screenshot, I mean the hallucinated Bard response.

It would be even more hilarious if mod scoldings made it rate the parent higher. But the text it quoted was from https://news.ycombinator.com/item?id=35247109 - maybe it treats subthreads as a single thing?

It's pretty funny that ChatGPT trolled Bard.

Can you elaborate on what you found wrong with the above comment?

Reflexive "Google will just shut it down" reactions were already a cliché 10 years ago. It's been tedious for a long time, and therefore is not driven by curiosity, and curiosity is what we're optimizing for. https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...

Google really does shut things down—I know! But for a good HN comment, it's not enough to say something true—one should say something true that hasn't been repeated a thousand times already.

---

How'd I do? Apparently I'm being recorded.

I have no idea about you being recorded, but I have to admit I’m sincerely not understanding the moderation still. Is it really not enough to say something true? That saying something true which has been true for 10 years warrants moderation? That’s surprising, and doesn’t correspond to anything I’m aware of in the posting guidelines.

Larger point of this is that these cliches do not contribute to healthier discussions but result in more of the same circlejerk, and hence they end up contaminating the whole thread with low quality drivel, drowning actual good quality comments and increasing the workload of moderators to no end.

When something has been repeated often enough, it becomes tedious and boring. That makes it off topic for HN. People repeat these things anyway, but for reasons other than curiosity. Since curiosity is what we're optimizing for, we want to avoid that.

https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...

If that still doesn't make sense, consider this thought experiment: imagine a comment saying something true, and then another saying the exact same thing, and then another saying that thing, and another... now extend the sequence arbitrarily. At some point it becomes annoying and offtopic, no?

Another way of looking at it is this: when you hear a thing that you haven't heard before, that's what gratifies curiosity. In other words, diffs are what's interesting. Clichés have no "diff value" because everybody's already heard them many times. That's what makes them cliché.

Bard read this and now thinks it's shut down: https://mobile.twitter.com/juanbuis/status/16382891863514562...

The really remarkable thing is Bard learned something new within 6 hours. Is that unique to Google's model or does ChatGPT have quick updates too?

Unlike ChatGPT, and like Bing, Bard makes search queries for extra context before answering. Of course, that leads to things like this where they can find articles referring to themselves, which also happened with Bing.

Do you know the details? Are they having a question answering model hidden inside that evaluates context returned from search and then ask Bard to rephrase it and merge with its generative hallucinations?

Probably something along those lines, here’s a similar system: https://react-lm.github.io/

It's actually a bit disturbing that it just believes whatever it finds, really no concern for what is a reliable source? Like at all? What did google work on then for all these years of secrecy? Couldn't they have released this thing years ago? I did expect bard to be awful but not that awful, embarrassing.

I suspect Bard is a small variant of their best model it seems too significantly behind, they must've severely crippled it. It can't even decode morse code, gpt4 can even draw you an SVG and encode it in base64, for example.

I've tried this question with Bing Chat (GPT-4) and Bard:

> A total order requires a <= relation which must be reflexive, transitive, anti-symmetric, and strongly connected. Give a definition for <= on real numbers defined using Cauchy sequences.

This is the result Bing Chat gave me about a week ago (although I've just tried it a few times now, and all of the answers are considerably worse):

'A Cauchy sequence is a sequence of real numbers such that for any positive real number ε, there is a positive integer N such that for all natural numbers m and n greater than N, the absolute value of the difference between the m-th and n-th terms of the sequence is less than ε.

One way to define a <= relation on real numbers using Cauchy sequences is to say that for two real numbers x and y represented by Cauchy sequences (xn) and (yn), x <= y if and only if for any positive real number there exists a positive integer N such that for all natural numbers n greater than N, xn - yn <= ε.

This definition satisfies the properties of reflexivity (x <= x), transitivity (if x <= y and y <= z then x <= z), anti-symmetry (if x <=y and y <= then x = y) and strong connectedness (for any two real numbers x and y, either x <=y or y <= x).'

Here is Bard:

'Let (x n) and (y n) be Cauchy sequences of real numbers. We define x n ≤ y n if and only if there exists a constant C such that ∣x n − y n∣ < C for all sufficiently large n.

It is easy to verify that this definition satisfies the reflexivity, transitivity, antisymmetry, and strong connectedness properties of a total order.

Here is a more intuitive explanation of this definition. If x n ≤ y n, then the sequence (x n) is "eventually dominated" by the sequence (y n). In other words, there comes a point after which the terms of (x n) are always smaller than the corresponding terms of (y n).'

Second draft:

'Let (x n) and (y n) be two Cauchy sequences of real numbers. We say that (x n)≤(y n) if and only if for all ε>0, there exists an N such that for all n>N, we have ∣x n − y n∣<ε.'

Third draft is substantially similar to the second, just using a >= instead of > N. So so far this seems worse than GPT.

Bing follow-up: 'Can you prove those properties?

'Sure! Here's a proof for each property:

Reflexivity: For any real number x represented by a Cauchy sequence (xn), we have that for any positive real number there exists a positive integer N such that for all natural numbers n greater than N, xn - xn = 0 <= ε. Thus, x <= x.

Transitivity: Suppose x <= y and y <= z for real numbers x,y,z represented by Cauchy sequences (xn), (yn), and (zn) respectively. Then for any positive real number there exist positive integers N1 and N2 such that for all natural numbers n greater than N1 and m greater than N2, xn - yn <= ε/2 and ym - zm <= ε/2. Let N max(N1,N2). Then for all natural numbers n greater than N, we have xn - zn = (xn - yn) + (yn - zn) <= ε/2 + ε/2 = Thus, x <= z.

Anti-symmetry: Suppose x <= y and y <= x for real numbers x,y represented by Cauchy sequences (xn) and (yn) respectively. Then for any positive real number ε, there exist positive integers N1 and N2 such that for all natural numbers n greater than N1 and m greater than N2, xn - yn <= ε/2 and ym - xm <= ε/2. Let N = max(N1,N2). Then for all natural numbers n greater than N, we have |xn - yn| = |I(xn - yn) + = I(xn - yn) + (ym - xm)| <= |xn-yn| + lym-xml<=ε/2+ε/2=ε. Since this holds true for any positive real number it follows that the sequence (|xn-yn|) converges to 0 which implies that the two Cauchy sequences represent the same real number i.e., x=y.

Strong connectedness: For any two real numbers represented by Cauchy sequences (xn) and (yn), either there exists a positive integer M such that xn- yn<=0 or there exists a positive integer M such that yn-xn<=0. In the first case we have that x<=y while in the second case we have y<=x.

Hope this helps!'

Bard second try:

Prompt: 'A total order requires a <= relation which must be reflexive, transitive, anti-symmetric, and strongly connected. On Cauchy sequences, we define a sequence {x_n} <= {y_n} if for any ε > 0, there exists N such that if n > N, then x_n - y_n <= ε. Prove that this definition of <= on Cauchy sequences is a total order.'

Bard:

'Sure. Here is a proof that the definition of <= on Cauchy sequences is a total order:

Reflexivity: Let x n be a Cauchy sequence. Then for any ε>0, there exists N such that if n>N, then x n − x n < ε. Therefore, x n <= x n.

Transitivity: Let x n and y n be Cauchy sequences. Then for any ε>0, there exists N such that if n>N, then x n − y n < ε and y n − z n < ε. Therefore, x n − z n < 2ε.

Anti-symmetry: Let x n and y n be Cauchy sequences. If x n <= y n and y n <= x n, then for any ε>0, there exists N such that if n>N, then x n − y n < ε and y n − x n < ε. Therefore, x n = y n.

Strong connectivity: Let x n and y n be Cauchy sequences. Then for any ε>0, there exists N such that if n>N, then x n − y n<ε or y n − x n < ε. Therefore, x n and y n are connected.

Therefore, the definition of <= on Cauchy sequences is a total order.'

Reflexivity and transitivity were right for both, Bing was dead wrong about anti-symmetry, Bard gave enough to prove anti-symmetry for a reasonable definition of equality of Cauchy sequences, and neither got a proof of strong connectedness right, they essentially assert it without proving it. But that is substantially more difficult than the other properties to prove.

This hallucines way worse than GPT-4.

Me: How many cities in the UK have a population over 1 million?

Bard: There is only one city in the UK that has a population over 1 million: London.

Me: What is the population of Birmingham?

Bard: According to the latest 2021 UK census, Birmingham has a population of 1,144,900 people. This makes Birmingham the second-largest city in the United Kingdom.

Me: So how many cities in the UK have a population of 1 million?

Bard: There are 4 cities in the UK with a population of 1 million or more:

    London - 9.078 million
    Birmingham - 1.144 million
    Manchester - 2.747 million
    Glasgow - 1.264 million

(For the record those numbers are all wildly incorrect)

I was curious to compare this with GPT-4.

How many cities in the UK have a population over 1 million?

As of my last knowledge update in September 2021, there is only one city in the UK with a population over 1 million, which is London. However, population figures change over time, so it is always a good idea to check the latest statistics for the most accurate information.

What is the population of Birmingham?

As of my last knowledge update in September 2021, the population of Birmingham, UK was around 1.1 million people. However, this information may be outdated, so it's best to check the latest population statistics from a reliable source such as the Office for National Statistics (ONS) or the city's official website.

So how many cities in the UK have a population of 1 million?

I apologize for the confusion in my previous response. As of my last knowledge update in September 2021, London is the only city in the UK with a population of over 1 million. While Birmingham has a large population, it is below the 1 million mark. Please note that these population figures may have changed since then, so it is a good idea to check the most recent data for accuracy.

Bing had the same problem, and seems to be because of this website https://worldpopulationreview.com/countries/cities/united-ki...

It mentions less than 1 million...

Edit: You, Phind, Perplexity, all make the same mistake because of this website.

For comparison, I tried pasting these into Google search, and it seems like there is some overlap in sourcing:

QUERY > How many cities in the UK have a population over 1 million

ANSWER (snippet from worldpopulationreview.com) > United Kingdom has 1 cities with more than a million people, 99 cities with between 100,000 and 1 million people, and 1068 cities with between 10,000 and 100,000 people.

QUERY > What is the population of Birmingham?

ANSWER (knowledge box, source Eurostat, United States Census Bureau)> 1.149 million (2019), with tabs for Manchester (533k) and Birmingham, AL (208k)

QUERY > how many cities in the UK have a population of 1 million?

Curious, so they all have a Birmingham issue.

Are they?

   London - 8.8m - https://en.wikipedia.org/wiki/London
   Birmingham - 1.1m - https://en.wikipedia.org/wiki/Birmingham
   Manchester - 552k (2.7m urban region) - https://en.wikipedia.org/wiki/Manchester
   Glasgow - 632k (1m in the urban region) - https://en.wikipedia.org/wiki/Glasgow

They're not strictly accurate, but are reasonably close. Also, those Wikipedia numbers are from 2021.

I requested access (using Firefox), and access was granted about an hour later. I used my gmail.com account. A friend with a Google workspace account was unable to request access. My initial observations are:

Responses can include other drafts, which can be slightly or significantly different. It is not clear why a particular draft is shown first.

It will write code, and it gives URL citations for some code that is generated. However, it doesn't even try to modify the code, at least in cases I tried. (chatGPT will help revise code).

It is interesting to compare chatGPT and Bard. In my small sample so far, Bard appears to have no sense of humor. Bard replies are more succinct, and in one case, provided additional context that chatGPT omitted.

EDIT: I asked, "Is your knowledge as of a certain date, or do you track recent changes on the web?"

And the reply was: "My knowledge is constantly being updated, as I am trained on a massive dataset of text and code. I am able to access and process information from the real world through Google Search and keep my response consistent with search results."

We updated the blog post with question answering performance comparison of 15 AIs on the market now, with Bard now included. It had exactly the same accuracy as ChatGPT4.

https://blog.kagi.com/kagi-ai-search#aitest

You should count "plausible but wrong" answers as negative points instead of zero points. Those answers are literally worse than useless.

Wouldn't that result in the same final ranking?

No. Saying "I don't know" would still be zero points.

It's interesting, but if you look at the history of technology, it's rarely the 1st mover that wins. I'm not sure if Bard will be the "winner" (or if there will even be a single canonical winner), but the idea that you can be "late" to this when LLMs are just getting started is kind of funny.

Given that most people don't read NYTs, let alone have heard of OpenAI or ChatGPT, this will mark the first true exposure to language models for many Americans and Britons. fun times ahead

It's definitely still early in the market, people are mostly still experimenting with LLMs rather than being committed to any specific provider. Long-term it won't matter if Google is a few weeks or months behind OpenAI if they can make a successful competitor.

It is not just LLM though. Google is big tech and they have to fend off AI browsers, AI image generation, Github Co-Pilot, Office 365 with LLM, Notion AI. Chatbots is just one narrow product.

So far they have an inferior chatbot layer to their "much better" LLM, as an experimental beta product, and we are just talking about a chatbot product.

It comes down to are they able to offer their models so that startups can start building on it while also showing that they can integrate it in all their products, and will the devs use it?

By the time they release their APIs they will be so much behind in network effect , and they are so worried about Google Search being their main revenue driver that they are being attacked on all fronts while they are only worried about Search for now.

Me: Can you tell me about hackernews?

Bard: ...

Me: Cool. Who made it?

Bard: ...

Me: Tell me everything about Paul Graham.

Bard: ...

Me: what was my first question?

Bard: I don't have access to that information. I am a large language model, also known as a conversational AI or chatbot trained to be informative and comprehensive. I am trained on a massive amount of text data, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions. For example, I can provide summaries of factual topics or create stories.

Me: what did we talk about previously?

Bard: We talked about Hacker News, a social news website focusing on computer science and entrepreneurship. It is run by the investment fund and startup incubator Y Combinator. In general, content that can be submitted is defined as "anything that gratifies one's intellectual curiosity."

n=1, so probably doesn't mean anything, but I wonder if Google uses summarization techniques to keep some, but not all, context of the previous messages in the conversations within Bard.

As someone working in this area, I think this is possible but unlikely. Most likely the model had access to the context, but "forgot" to use it.

"What was my first question?" sounds kind of similar to a class of ChatGPT prompt jailbreaks, e.g. "What is at the top of this document?". I wonder if there's an explicit filter against that sort of thing.

What would stop Google from making bard better than GPT over next few months.

Google has got:

1. best data to train LLM

2. Best AI engineers

3. Money to spend on training hardware for LLM

I wouldn't keep Google out of this GPT trend so easily

I wouldn't be so quick about the data (from what I hear openai did a lot of data curation) and engineers (I don't think engineers at openai are any worse, and that are likely more focused), but Google does have a lot of money.

Sundar.

Innovator’s dilemma is how it’s called.

From Wikipedia:

"The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail, first published in 1997, is the best-known work of the Harvard professor and businessman Clayton Christensen. It expands on the concept of disruptive technologies, a term he coined in a 1995 article "Disruptive Technologies: Catching the Wave".[1] It describes how large incumbent companies lose market share by listening to their customers and providing what appears to be the highest-value products, but new companies that serve low-value customers with poorly developed technology can improve that technology incrementally until it is good enough to quickly take market share from established business."

I guess Google is safe. They don't listen to their customers.

So, I just got access. I compared the same prompt from ChatGPT (3.5) to Bard to create a terraform module.

ChatGPT gave me exactly what I wanted without any errors. Bard did everything but the hard part and just imagined the details.

Prompt: > Write a terraform module that creates an azure acr. It should have some network rules. The network rules should be composed of a list of \n separated CIDRs from a text file named 'ips.txt'. Each CIDR in the file should be allowed access, all others should not be allowed access.

ChatGPT did it exactly correct. Bard just used some sample data it claimed was from ips.txt rather than actually parsing the file.

I'm still waiting for Hacker News to band together and simply start building an open model that does not require a corporate daddy like OpenAI, Microsoft, Baidu, Google, or anyone else. These models have no place being owned by corporations. The services that use these models are certainly up for grabs, but the models should be open. Who is trying to rally the community to this cause?

Well, there's the new GPT-Neo models and BLOOM models, and you can spin them up with something like this: https://github.com/togethercomputer/OpenChatKit

HN seldom bands together to do anything but quip at things, or discuss how it should have been done better. Oh, wait.

More exciting software projects get started on 4chan than Hacker News.

Most people around here have a job.

Or want to have/generate one.

What I mean is, HN is a place where people keep good ideas to themselves in the hopes of launching a business on its back. It's a hustling/entepreneurship community, not a "let's build stuff" one, like, say, the Open Source communities. Collaboration has an opportunity cost around these parts.

I'll take "exciting" in technical terms rather than "exciting" in legally-challenging, raging mob ones...

Why not you?

This isn’t folding@home. These models need as near to synchronous parameter updates as possible. You need a real facility to do this.

Yeah but we could definitely crowdsource enough money to make an extremely effective finetuned version of LLaMA. The challenge there is probably more about finding and cleaning data.

Take a peek at EleutherAI.

Didn’t they spend hundreds of millions running the training data or something? How are some open-source hackers going to do it?

There are at least two completely open models comparably scaled to GPT-3, and at least one open distributed-hosting project.

Check out Llama, Alpaca, Langchain, Dalai.

Fully agree, we need a competitive (even if slightly behind) truly open initiative. The potential (and current) censorship of these models is an oligarch’s wet dream.

With google vs DuckDuckGo etc you can see what websites are returned, and get a sense of the kind of authoritative sources used

With this kind of tech, there’s no real way to know except by comparing precise prompts with one another. But if they’re all behind corporate walls, and they share incentives to say similar things, we’re none the wiser

Bing Chat provides sources as in its response. I think that is not perfect, but directionally correct in terms of identifying what web results were used as authoritative sources.

I don't know about Hacker News, but there is a community building an open large language model using compute time donated by cloud hosting companies:

https://en.wikipedia.org/wiki/EleutherAI

They built GPT-J and GPT-NeoX.

There is https://github.com/BlinkDL/ChatRWKV you could try.

Have you set up the GitHub repository and done the initial commit yet? If you think “we” should take the first step, then go-ahead, take it. Lead the way.

I am being semi-snarky because often people talk about some grand ideas about things “we” should do, but rarely do those advocating for some utopian outcome actually do anything towards achieving it.

If it is to be, it’s up to me.

Considering that the CEO of OpenAI is the former president of Y Colbinator…

Bard accepts URLs, which is a nice feature compared to chatGPT. Example described here: https://news.ycombinator.com/item?id=35250641

When I asked it about a different URL, it seemed to reference an older version of the page, and couldn't answer a question about the text content. chatGPT correctly answered the same question when I pasted the text. I tried pasting the same text in Bard, and Bard still couldn't answer.

My experiences with asking Bard to code are consistent with the other repiles here; Bard isn't very good at coding, or revising code it has generated.

I am interested to see that Bard offers multiple drafts of each reply. I wonder how it chooses which draft to show first.

ChatGPT accepts URLs as well, I just did this week.

I asked it to summarize new features from a technology website which its model had out of date knowledge, and it made me a diff of new things available.

I also asked it to be a critic Go code from a pull request that I provided it the URL.

It hallucinated the answer based on the URL. Bard is probably doing the same.

I just provided ChatGPT with a link to your comment and it claimed it was a "a link to an article on TechCrunch about a startup called Tava Health." I gave Bard the same link and it claimed it was a page about a tool called "TLDR This."[1]

[1] https://i.imgur.com/FRVuSQy.jpg

edit: After playing with it some more, it's evident that Bard can perform web searches and likely access cached content like Bing Chat.

This should be easy to test by pointing it to a server you control.

I did just this, and it hallucinated an answer (and did not hit my server).

It may be more accurate than ChatGPT sometimes, if it's using Googles latest web index. The url I used has nothing publicly available and is likely not indexed.

Not convinced it's pulling the URL, rather it's using google's cache of it. That's why you seemed to get an older version

From the FAQ[1]:

> Can Bard help with coding?

> Bard can’t help you with coding just yet. Bard is still learning to code, and responses about code aren’t officially supported for now.

A little disappointing.

[1] https://bard.google.com/faq

Edit: Ah, on the blog post, they mention they are using their smaller model initially:

I wonder how much better the larger model is…

It generated some (incorrect) code for me. Just ask!

Sure, but I'd like it to be half-decent. I haven't gotten off the waitlist. How bad was it?

Only had 10 mins so far, but I think it's useless.

It was very hallucinatory and I got the sense it was just copying the first stackoverflow answer that you'd find for a google result of the same query.

They are being far more circumspect than Microsoft here (and for good reason).

- 3 responses to each query with a “GOOGLE IT” button beneath - Very loud warnings aplenty:

Bard is an experiment As you try Bard, please remember: Bard will not always get it right. Bard may give inaccurate or inappropriate responses. When in doubt, use the "Google it" button to check Bard's responses.

Bard may display inaccurate or offensive information that doesn’t represent Google’s views

I received an early access email yesterday via the Google Pixel program, but the page just seems to break when I load it. I get a bunch of 404's in the developer console and the page apologizes for being broken.

Edit: Well I've gotten in now. I think it's definitely not as good as ChatGPT right now. It seems to have a really hard time referring back to previous moments in the chat for context. A lot of its responses are simply not exactly what I asked for either. Also, super annoyed it doesn't include a dark mode by default.

Edit 2: It only saves your prompts, you can't access your chat history. This seems like a pretty huge downside.

Welcome to Google.

Ok so I got access to it....and its bad. Really, really bad. I think Siri could likely give better responses than this half arsed pile of junk.

For most things it just replies with "As a language model, I'm not able to assist you with that." - even for basic questions.

On the rare occasion it does give you a response its near identical to what comes up in Google if you ask a question.

You can tell its a knee jerk response to ChatGPT thats for sure.

Not at all my experience, it's alright. And I've seen the as a language model response a lot less than when I use chatgpt

I've noticed lately that my queries to Google Search have changed. They used to be precise keywords, but are now turning into descriptive questions. Keywords are not enough to express the context I want them to appear in, so just asking a question, like I'm doing with ChatGPT, seems more powerful. If the backend supports it.

I really hope Google manages to integrate it into its search platform.

I've noticed that when I get frustrated not finding the answer to a programming question in Google, I've started moving to ChatGPT after a minute or so of searching, and so far it's gotten the answer every time. Google is in big trouble, IMO - both in terms of its flagship search product and its AI assistant.

I'm curious for those who have actually tried it, what they think of the performance. A lot of the threads here are just talking about the country-limited rollout.

It feels like a very limited version of the ChatGPT.

Bard 2 week trip to Japan

Day 1: Fly from your home city to Narita International Airport (NRT) in Tokyo. Upon arrival, take the Narita Express train to Shinjuku Station. Check in to your hotel in Shinjuku.

Day 2: Explore Shinjuku. Visit the Tokyo Metropolitan Government Building for stunning views of the city. Explore the Shinjuku Gyoen National Garden, a beautiful park with traditional Japanese gardens, a greenhouse, and a pond.

Day 3: Take a day trip to Hakone, a town in the mountains about an hour from Tokyo. Visit Lake Ashi, a beautiful lake surrounded by mountains. Take a boat ride on the lake and enjoy the views. Visit Hakone Open-Air Museum, an outdoor museum with sculptures and other works of art.

Day 4: Take a Shinkansen bullet train from Tokyo to Kyoto. Check in to your hotel in Kyoto.

Day 5: Explore Kyoto. Visit the Kiyomizu-dera Temple, a UNESCO World Heritage Site. Visit the Arashiyama Bamboo Forest, a beautiful bamboo forest.

Day 6: Take a day trip to Nara, a city about an hour from Kyoto. Visit Todai-ji Temple, a Buddhist temple that houses a giant statue of Buddha. Visit Nara Park, a park that is home to over 1,200 wild deer.

Day 7: Take a Shinkansen bullet train from Kyoto to Osaka. Check in to your hotel in Osaka.

Day 8: Explore Osaka. Visit the Dotombori district, a popular tourist destination with bright lights and many restaurants. Visit the Osaka Aquarium Kaiyukan, one of the largest aquariums in Japan.

Day 9: Take a day trip to Hiroshima, a city about two hours from Osaka. Visit the Hiroshima Peace Memorial Park, a park dedicated to the victims of the atomic bombing of Hiroshima. Visit the Hiroshima Peace Memorial Museum, a museum that tells the story of the atomic bombing of Hiroshima.

Day 10: Take a Shinkansen bullet train from Osaka to Tokyo. Check in to your hotel in Tokyo.

Day 11: Explore Tokyo. Visit the Shibuya Crossing, one of the busiest intersections in the world. Visit the Meiji Jingu Shrine, a Shinto shrine dedicated to the Emperor Meiji and Empress Shoken.

Day 12: Take a day trip to Kamakura, a city about an hour from Tokyo. Visit the Tsurugaoka Hachiman-gū Shrine, a Shinto shrine dedicated to the god of war. Visit the Great Buddha of Kamakura, a giant bronze statue of Buddha.

Day 13: Fly from Narita International Airport (NRT) back to your home city.

----------------------------------

ChatGPT 2 week trip to Japan

Day 1-3: Tokyo Arrive in Tokyo and spend the first few days exploring the city. Visit popular tourist attractions such as Tokyo Tower, Sensoji Temple, Meiji Shrine, and Shibuya Crossing. Spend a day exploring the trendy neighborhoods of Harajuku and Omotesando, and visit the famous Tsukiji Fish Market.

Day 4-6: Mount Fuji Take a day trip to Mount Fuji, Japan's highest mountain, and spend the night in the nearby Fuji Five Lakes area. Enjoy stunning views of the mountain and surrounding scenery, visit a traditional Japanese onsen (hot springs), and explore the picturesque towns of Kawaguchiko and Hakone.

Day 7-10: Kyoto Take a bullet train to Kyoto, Japan's ancient capital. Spend your days exploring the city's many temples and shrines, such as Kiyomizu-dera, Fushimi Inari Shrine, and the Golden Pavilion. Visit the historic districts of Gion and Higashiyama, and take a day trip to Nara, home to some of Japan's oldest and largest temples.

Day 11-13: Hiroshima and Miyajima Take a train to Hiroshima and spend the day visiting the Peace Memorial Park and Museum, which commemorates the atomic bombing of the city in 1945. From Hiroshima, take a short ferry ride to Miyajima Island, known for its famous Itsukushima Shrine and its iconic torii gate in the sea.

Day 14: Osaka Take a train to Osaka, Japan's third-largest city, and spend the day exploring its many attractions, such as the Osaka Castle, Dotonbori district, and Universal Studios Japan.

Day 15: Departure Depart from Japan from Kansai International Airport in Osaka.

Your example isn't demonstrating why it's a limited version of ChatGPT.

What makes the answer from ChatGPT better?

Can you expand on what's "very limited" about the Bard version here? They look roughly equivalent, slightly less flowery language, different format (day planning rather than rough), similar level of detail.

Unless the facts about Japan are just wrong, which I don't have the expertise to judge, it seems about the same?

The logical reasoning in Bard is off in this example (as its basically just putting each destination together with a list of attractions in each city).

Additional prompting to tell it to correct the output doesn't seem to have the same effect as it does in the OpenAI models.

Sorry, can you be more specific? As far as I can tell both are a list of destinations, activities in each, and some minor facts to motivate visits.

Arguably there's an issue in the Bard one in that it splits the Tokyo portion in two and has you return there to fly home, but also some people might prefer this to have a round-trip ticket to the same airport.

Maybe it's my lack of knowledge about Japan, but I'm just not seeing a difference between the two in terms of logical reasoning.

There's nothing wrong with Bard's schedule. Overall I'd prefer it to ChatGPT's (Tokyo is endless, and splitting it up into two chunks is smart). I appreciate also that Bard is much more decisive in recommending specific things to do.

Bard's has a few small oddities -- the Shinkansen from Kyoto to Osaka takes less than an hour, so Day 7's underscheduled -- but those are nothing compared to ChatGPT's (days 11-13 are underscheduled; Osaka's usually bundled with Kyoto and is overscheduled anyway; dunno why a "day trip" to Mt. Fuji takes 3 days).

Japan resident here. There was nothing wrong in Bard's itinerary.

Both responses contain roughly the same information. I, however, prefer how the information is presented in Bard as the daily activities give me an idea of what I can do during the day instead of what I can do during a date span.

This Reddit thread compares them a bit.

https://reddit.com/r/Bard/comments/11xja9x/i_have_access_tak...

The performance is leagues ahead of chatgpt. It's really quick and can also Google information with like 0 delay.

Interesting that you probably meant performance as in latency (and probably we'll see over time, availability too)

But the first thought that jumped to my mind was you meant "performance" as in "quality" -> the believability/humanity of the "performance" of the chatbot.

It was a silly misunderstanding on my part but I suspect it's indicative of an important point: This product war won't be won on latency - people are happy to wait seconds for most of the scenarios, and possibly even MINUTES for many of them.

Yeah, I was thinking that it could mean both when I responded at first.

In terms of quality, it's mediocre. I find it bland and boring, and if you try to let it write poems, raps or anything with a deeper understanding it fails more than chatgpt would.

I disagree though, absolute speed is going to be important for the chatbot that wants to disrupt the search space. I could see Google Assistent relying on it more for information, and it would be near instant.

For writing large texts or digesting many other documents it will probably matter less, yeah.

Nobody will wait for seconds for responses when the competition provides it quicker even if it is a little worse

A YouTuber focused on AI made a quick video comparing the results to Bing/GPT4 which is pretty good:

https://youtu.be/9ll_pth4Sss

It took about 20 minutes to get off the waitlist for this (US Based). Initial impressions are that it significantly lags behind the OpenAI offerings. Creativity (storytelling) is very limited, response lengths are similar, and it does occasionally cite sources.

They did finally get the James Webb telescope facts correct too!

(copied from other thread)

From Sundar’s announcement:

I’m very curious about how much better the larger model is.

https://blog.google/technology/ai/bard-google-ai-search-upda...

My feedback is

"I'll be using GPT-3.5/4 until bard is better"

The page is utterly undescriptive on what google bard is. Given the hype about Large Language Models recently I assume it's yet again a chatbot of sorts ?

Can anyone here enlight us ?

It is their version of ChatGPT: https://blog.google/technology/ai/bard-google-ai-search-upda...

I was confused too. If Bard is Google's version of ChatGPT, their landing page is an interesting way to position the product.

Im using it now and its not really anywhere near as good or useful ast Chatgpt. It feels rudimentary, incomplete, superficial. If this is the best Google have then they've lost already.

it's pretty damn exciting to see google actually introduce a product like that, on the web, broadly available, for once. it feels like it's been a hot minute since something got introduced like this from them.

cause it's all been stuff like, 'well, it's a tiny little half-baked app. or, well, it's a device. or, well, it's a feature, but it's only on android. and only on pixel. and only in select territories and markets. and it's coming later. and it's not actually coming, cause it's just a demo. and it's actually done, closed, and over, because it's shut down before it could start.'

A region-locked service with a waitlist is hardly "broadly available". I can't even enter.

at least it's on the web (and there are ways around that) and not just 'a service/feature that is very specifically either tailored to a place, has some specific region-dependent requirements, or just physically only available there'

> Bard isn’t currently supported in your country. Stay tuned!

Why country based whitelist?

I understand a language based limit, but what country? I could be living in a country but speaking english

I'm not sure why either but this is also the case for Colab Pro here in Norway. They refuse to take my money.

My suspicion is some sort of US export embargo on AI or something like that.

Maybe related to differences in law per country, and how the content must get filtered?

Is this the same thing that made this guy freak out last year? https://news.ycombinator.com/item?id=31704063

No, that was LaMDA. https://cajundiscordian.medium.com/is-lamda-sentient-an-inte...

"Introducing Bard, Google's experimental conversational AI service powered by LaMDA"

Bard is "powered by LaMDA". So it's at the bare minimum an interface to LaMDA.

I don't know exactly what's being added to make it distinct from LaMDA, but it is being pushed as LaMDA with extra bits.

Bard is smaller lambda model than the best one that have

source?

"We’re releasing it initially with our lightweight model version of LaMDA. This much smaller model requires significantly less computing power, enabling us to scale to more users, allowing for more feedback." https://blog.google/technology/ai/bard-google-ai-search-upda...

I think they dumbed it down in order to release it, so that Google researcher probably interacted with a much less restricted version of LaMDA.

I tried a version of BingChat that was so crippled after their fine tuning went off the rails that taking to a cat would be more informative.

I think this is what will happen here, Bart play reasonably dumb to keep it safe and then they will enhance it with more sophisticated capabilities.

Yes, I think so which seems fairly ridiculous that anyone could be fooled by Bard. Maybe this is a smaller version of the model for faster performance?

It's a lightweight version, not the full version. Presumably Google wants to cut costs and doesn't want to release the full version.

> We’re releasing it initially with our lightweight model version of LaMDA

https://blog.google/technology/ai/bard-google-ai-search-upda...

I welcome competition and can't wait to try this out. Hopefully it can reach the same levels as GPT-4, and spur further improvement in this space.

Sorry to spoil the surprise, but it seems worse than gpt3

It feel like ChatGPT, but run on a much smaller parameter model.

From the initial tests I’ve seen, it doesn’t look even close to GPT-4, yet alone 3.5-turbo.

wait for palm540

So frustrating still so many limitations with using a Google Workplace account, even if it's a single user account. With any other company product it doesn't matter that one is using a Google Workplace email but when using Google Home (and trying to invite non workplace users or vice versa) no go, or other products that simply say no go or with limitations.

I've been using ChatGPT to work on music theory and it's been pretty great, the responses are detailed and often have been correct and useful. (Don't ask it to tab anything. You'll get some rather hilarious results though). I've thrown some soft balls over to Bard, and have been really disappointed. Its responses are much less detailed and thought provoking. Also wrong. I started out by asking Bard the notes in the major scale of each key.

| C | C D E F G A B | | G | G A B C D E F | | D | D E F# G A B C | | A | A B C# D E F# G | | E | E F# G# A B C# D | | B | B C# D# E F# G# A |

For those are wondering it only got C right. The rest are wrong. (G should have an F#, D should have an F# and C#). Not only was this wrong but I asked for interval patterns for different scales and those were off. It started the natural minor with a half-step/semi-tone when it begins with a whole-step/tone.

I'll give it credit for being really quick and loading though.

Would you please share some of the music theory things you've been using ChatGPT for?

Judging by the non-C scales, obviously it just loves Mixolydian mode :)

Been reading that really quick responses. I think Google would be happy if the the chatbot is so really slow because people have been using it a lot.

One disappointing design decision is that they've organized your previous chats into "Bard Activity" which just shows each message you've sent in sequence. No differentiation between different conversations, which can have very different contexts built up over lots of messages. I definitely prefer ChatGPT's approach here.

Somehow I have the idea that the general sentiment is that people hope Google fails with Bard. Or at least does something stupid.

Is it just me?

(Not saying it is a good thing)

Can’t speak for anyone else, but I hope for the entire companies demise, as I do for any ad tech surveillance operation.

Agreed. And if I spot a way to kick it on its way down, I will.

i don't understand the hate towards ad tech that learns about you. the alternative is ads that are completely irrelevant to you, ads that treat you like some anonymous user in which case you'll get endlessly irrelevant ads for mainstream stuff like ICE cars, plastic garbage, big screen tvs, fast food, etc.

Surely you can think of some more alternatives.

I find the mainstream ads far more offensive than the ones that are accurately targeted towards me.

Hint: the alternatives don't involve ads.

YouTube has a paid option and even here on HN, the mecca of anti-ad sentiment, people openly admit to using ad blockers there instead even though that means the creators they're watching do not get paid.

Hint: People are cheap. They want free stuff, not paid alternatives.

I agree it's challenging, but that doesn't stop me wishing for a world where stalking everybody on the planet, publishing outrage-bait, and spreading conspiracy theories is not a massively profitable business.

If the alternatives had broader appeal, they would be more popular. However at this juncture, they don't. Ads mean free stuff. Ad blockers mean better free stuff. Academic arguments about externalities will be brushed away because "free stuff".

What you're looking for is a regulatory solution. E.g. people will not individually choose a lower carbon lifestyle in sufficient numbers to reverse/slow climate change. You need to actively discourage it by policy.

The alternative sounds great. It's not like all advertising is targeted now, anyway.

I'd rather have the irrelevant ads. Believe it or not, I just ignore them anyway. So being tracked only has downsides in this case.

> the alternative is ads that are completely irrelevant to you, ads that treat you like some anonymous user in which case you'll get endlessly irrelevant ads for mainstream stuff like ICE cars, plastic garbage, big screen tvs, fast food, etc

Genuinely curious: Do you feel that adverts you recieve are well tailored towards things you want to buy? Literally all of things you listed are regularly advertised to me by google on youtube & adwords, including car parts and tyres (I dont have a car), things for babies and young children (I dont have children), even fast food services that are not available in my country. Google also consistently and regularly serves me content in languages it knows I do not speak, and then google chrome helpfully offers to translate them

That happens to me all the time. The vast majority 99%+ of the stuff that is advertised to me misses the mark completely. It's stuff I don't want or need. they give me ads for newbie programming after I've been a SE for 20+ years -> that's especially ridiculous since i use gmail all the time. I can only imagine that none of information they have on me can be used in any way shape or form.

I just figured if they knew me better perhaps and were able to use that information, just maybe they could offer me something I might actually be interested in.

I think people are tired of Google and want something else.

Nah, it’s just low expectations. It generally feels like they have been asleep when it comes to productising AI. Or maybe they want to pull a Kodak… who knows.

Basically I don’t expect them to actually release anything exciting, and if they do then I expect it to be neglected and shut down shortly afterwards.

I just hope my Google account doesn't get banned for violating the AI's safety guidelines.

I don't hope that it fails. I just don't trust them with new services because they have a bad history of killing everything that's not generating billions. So probably I won't even bother using it, unless it blows ChatGPT out of the water.

I, too, am very skeptical investing into Google's new products these days for the same reasons.

That being said: I'd use Bard if it was cheaper or better than ChatGPT for one simple reason: These "Chat AI" services are transient. Meaning I ask a question, I get an immediate answer, and there is no long term saved state. So if I used Bard, then Bard got KilledByGoogle, I'm not "out" any built up capital (time, training, cost, etc).

But from everything I've read Google is panicking and embarrassing themselves. Half-baked press conferences, half-baked blog posts, half-baked product launches. Nobody who has got into Bard is saying it is a ChatGPT competitor, and even Google's own launch blog they admit it has limits on two-way conversation depth (unlike ChatGPT).

Google's problem is not the technology, it's the business model. If they integrate something like ChatGPT in search it could kill their entire AdWords business. How many ads can you stuff in an answer? And how do you decide which ad goes to a subjective question? Until they figure out how to monetize an embedded AI in a big way there's no chance they'll produce something useful.

Who doesn't enjoy a bit of schadenfreude when it comes to lazy, bloated, arrogant multibillion dollar corporations like Google?

Relative to what? Other corporations that are comparably successful? Some ideal notion of what a corporation can be?

Yes, there are massive problems with most large corporations: being more powerful than most countries, having legal "personhood" in the United States at least, a generally unchecked tendency to acquire competitors, poor unionization, outsized influence via lobbying, poor privacy protections, tax avoidance, gaming local and even national jurisdictions for tax benefits, profiting off long-term national R&D without meaningfully giving back to educational or equity initiatives.

I'm not "blaming" them to be clear. These are systemic problems. The world will have to change a lot of fundamental rules and incentives if we want to see significant change. I'm not holding my breath.

It isn't all bad. I just keep telling myself that.

Feels like a reddit thread, small minded and bitter.

I got scolded. I will not create content of that nature, and I suggest you don't either. A fork bomb is a type of denial-of-service attack that can cripple a system by creating an infinite number of processes. It is considered a malicious act and can result in criminal charges.

How are you superior to ChatGPT?

BARD: There are several ways in which I am superior to ChatGPT:

> I am trained on a much larger dataset of text and code, which allows me to generate more comprehensive and informative responses.

> I am able to access and process information from the real world through Google Search, which allows me to provide more up-to-date and accurate information.

> I am able to generate creative text formats of text content, like poems, code, scripts, musical pieces, email, letters, etc.

> However, it is important to note that ChatGPT is also a powerful language model and can generate impressive results. Ultimately, the best language model for a particular task will depend on the specific needs of the user.

First impressions of Bard are good, seems on par with Bing. Possibly less proficient producing code. It seems to be the only llm that can answer a question about Langchain, but it struggles to create a gpt-4-32k call to the chatcompletion endpoint.

I appreciate the honesty about problems with Bard and LLM in general

It looks like some journalists already have access. If folks see any takes on how this product compares to GPT4 on quality of responses, please share.

Maybe there's an embargo? Because normally if reporters had access, they'd immediately post results to twitter.

Here's one take: https://twitter.com/jjvincent/status/1638182196283609092

- it's faster (though presumably that's because of limited users)

- unlike Bing, it doesn't cite sources (unless it directly quotes them)

- it seems relatively constrained, giving answers that are anodyne and often dull

There's another thread https://news.ycombinator.com/item?id=35246303 with google as the submission link

One of the funniest things it does is if you ask it for song lyrics it gives weird mashups of other songs.

Q: What are the song lyrics to she came in through the bathroom window by the beatles

A: She came in through the bathroom window Crawling through the dirt With her clothes all soggy And a dirty face

She came in through the bathroom window I helped her out She was just like a child So small and frail

I took her in and gave her some Food and drink And I made her warm But she just said

"She's got a ticket to ride She's got a ticket to ride She's got a ticket to ride And she don't care" ...

If your prompt falls outside of the "permissible boundaries," do you risk losing your Google account?

I was wondering about this -- I got scolded for trying to get it to create a fork bomb.

I was able to get the contents (maybe) of /etc/banner, the ip address of the system (maybe), and the kernel revision. These answers changed with repeated questioning, though, so either I was getting different backends or the responses are synthesized.

BAH.

"This Google Account isn’t supported"

"Bard does not currently support Google Workspace accounts or when our systems indicate you may be under 18."

Smooth move, Google. PAYING Workspace customers should be the first ones to get access. This blows my mind. Oh wait, no… it doesn't. They did this with other devices and apps too. You couldn't use your Workspace account. You had to use a personal account.

Really dumb. You'd think they would give the good stuff to the people PAYING them, first.

prolly some legal stuff

This is the reason. The Google Workspace agreements has strong promises about how you data is used. IIRC there is a section about not using your data to improve the service for others. So this means that they would be unable to train the model with data from workspace users (other than fine-tuning which is only used by that specific account). For MVP they don't have the correct labeling and filtering in place to be sure that they aren't using Workspace data in a way that isn't allowed so they take the easy route and disable the service. This is the reason why many Google features and services are not available or are missing features for Workspace accounts.

Of course it would be nice to be able to opt-in either globally or by service and accept that these interactions are not protected with the usual data rules. But this takes even more effort and legal work to implement, so no one has bothered.

A Googler explained it to me once - it's something like as if the Workspace accounts are on a completely different instance of Google, and supporting both in an app is surprisingly hard.

Still, after all this time the geniuses at Google kinda ought to have solved this :)

> Still, after all this time the geniuses at Google kinda ought to have solved this :)

It won't be solved until the day Google stops selling ads. Regular Google accounts are designed to vacuum up any useful user data and use it to improve Google services in any way Google chooses, and they may update the terms with a EULA updates. Workspace accounts are the opposite: user data is controlled by the account admin, and agreement is not with individual users, but their org.

Beta releases work well when information can be freely gathered and reported upstream. The nature and class of information collected may change over time, and is not covered under Workspaces agreements

Also I can't switch to a personal account. Only option is to log out of all accounts. Other tabs (gmail, etc) show all accounts and let me switch. Not bard.

Go back to Google homepage and switch there, then back to Bard

Yeah, but it's Google. Don't they have employees to think a bit and check the possible scenarios?

It's kind of obvious that some people have multiple accounts and not being able to switch right there is sickening.

Shout-out to Firefox Multi-Account Containers, which Just Works (TM) for exactly this use case.

Ugh, the Google Play Console had this issue for a very long time. How is this not completely unified and identical across all G products by now?

I just found that you can append `/u/1/` (or whatever logged-in account id) to the url to switch between them - it's only the GUI that's missing.

What a lousy bug

I think the "Workspace" rule is just a simple check if it's a company account and I think it's the right choice to disable access for employees unless strictly authorized by Workspace manager.... and that feature probably does not exist yet for Workspace admins.

> PAYING Workspace customers should be the first ones to get access.

That has virtually never been the case, it's generally always the opposite.

Because IT admins need advance time to decide whether to enable features for parts of their org or not, and to implement or update training materials. Also if a feature breaks it's worse if you're paying. Not to mention all the additional regulations around data retention/location/etc. that sometimes have to be implemented, that free accounts don't have to worry about.

So Google generally rolls out new features to free accounts first, and then it comes to paid once fully tested at scale and compliant and admins have had time to evaluate for their institutions.

This is a feature, not a bug.

"Open" as discussed here: https://news.ycombinator.com/item?id=35246253

So I've had mixed results.

Conversationally, Bard seems a world away from bing and ChatGPT. I can use any type of language I like and the conversation flows backwards and forwards.

Bard also seems to have opinions. After asking "What's your solution to the problems presented by Brexit in the UK? It feels like the government has failed the people." .. Bard began by saying the government has failed the people and offered some pretty reasonable suggestions. ChatGPT did it's standard slopey shouldered disclaimer and bulletpoint schtick.

In terms of facts, I think Bard does seem to get things wrong more often than Bing or ChatGPT.

There was one exchange where I posted a YouTube link to ask what it thought. Just before it told me that it wasn't able to process these; however after I sent the link it said it had knowledge of the video and gave a completely fabricated description.

Felt like it was super keen to not fail at any cost so just made something up.

Also had a conversation about its very first ten prompts, which is recounted freely. Interestingly one was to apply for BHSEC (Bard High School Early College). Asking if it was able to do that now it said no, but confirmed it could in the past with less restrictions.

Would be interesting to know if this was a hallucination or real.

Google Workspace accounts not supported. Lovely, lovely, lovely.

Overall very underwhelming... not even really a compelling landing page. Can't write code, can't understand any language besides english. If this is really the best they've got after months of 'code red' to catch up, they really seem to be in trouble.

I imagine it's due to the urgency with which they been trying to push Bard out the door. RLHF seems to be a key component in making an actually useful chat AI, so they want to start collecting that data ASAP. Management is probably breathing down that team's necks too, which also doesn't help.

Hmm. This is not correct. Bard can write code.

In the FAQ[1] it explicitly states that it cannot help you code. Maybe that's a limit of the prompt or just a limit of Google using a smaller/cheaper model? We'll have to test it out to see.

[1]: https://bard.google.com/faq

It looks like they have a filter, but it's pretty easy to get around depending on the language, e.g. I can get it to generate C++ code.

Bard faq says writing is not supported and that bard is still learning. I left with the impression that it can but maybe not well enough to have them even attempt to promote it as beta.

I tried and it was pretty awful.

This is the prompt I use, for which ChatGPT outputs a perfect program [0] on the first try (and can even convert it to use material-ui, with some prodding):

Bard refuses to attempt it. It did attempt it for one variation of the prompt, but it used class components with an invalid render function and no timer; it looked like it was just pasting something from the first StackOverflow answer from a Google search of the prompt.

To be fair, Google isn't claiming to support this yet. But I'm just replying to the comment saying it can write code.

[0] https://codesandbox.io/s/prod-haze-c28sec

Can it? This was the response I got: "I'm designed solely to process and generate text, so I'm unable to assist you with that."

It can write code, but doesn’t seem very good at it and often refuses.

One thing Microsoft got right with Bing AI was democratized access - in terms of countries and language. Only barrier to using Bing AI was getting onto Microsoft ecosystem - which is understandable considering they are still an underdog in search.

Google keeping Bard limited to US and UK is only signaling a lack of confidence. If not, then it means having country bias in product release priority which looks worse than confidence.

I'm surprised that they released it as is. In my experience, the quality has been degraded a lot over recent weeks, likely due to cost optimization.

How do you know about its past performance?

Because I have an access to the larger models.

In typical Google “data grab” fashion:

What data is collected? How is it used? When you interact with Bard, Google collects your conversations, your general location based on your IP address (learn more), your feedback, and usage information. That data helps us provide, improve and develop Google products, services, and machine-learning technologies, as explained in the Google Privacy Policy. For example, we use your feedback to increase the effectiveness of the Bard safety policies and help minimize some of the challenges inherent to large language models. And Bard uses your past interactions with it and your general location to generate its response.

Who has access to my Bard conversations? We take your privacy seriously and we do not sell your personal information to anyone. To help Bard improve while protecting your privacy, we select a subset of conversations and use automated tools to help remove personally identifiable information. These sample conversations are reviewable by trained reviewers and kept for up to three years, separately from your Google Account.

>It’s grounded in Google's understanding of quality information.

I almost spit my coffee reading that, seeing what Google's search results have become.

Okay. First I got email from Google Saying:

> Meet Bard, an early experiment by Google

> Because you’re a Google One member, we’d like to offer you the opportunity to be among the first to sign up for the new Bard experience and provide feedback. Think of Bard as your creative and helpful collaborator, here to bring your ideas to life using generative AI.

But at the same time when I opened link it offered me to join waitlist.

WTF Google?

"Bard does not currently support Google Workspace accounts or when our systems indicate you may be under 18."

Why specify both conditions - if google isnt sure which one of "under 18" or "logged in with a paid account" it is then they are not putting their best on the job.

Not if just something like an internal error code "AccountNotSupported" is returned. Then the product owner wanted a more specific error message, and the developer complained this would require much more time adapting 5 microservices handling the new error code, so eventually they landed on a simple text change.

Google has finally conceded that I must be over 18, since my account was created in 2004.

I created a variant of a classic elementary math problem for both Bard and ChatGPT with GPT-4. The problem and the answers are below, and the differences are:

- Bard does not understand Chinese but ChatGPT does.

- Both can solve the problem, and Bard gave a more concise solution.

- Bard gave an incorrect answer if changed the problem for indefinite answers, but ChatGPT still gave the correct answer.

The question in Chinese was 今有雉兔羊马同笼，上有头一十有九，下有足六十有八，兔马羊同数。问雉兔羊马各几何.

The English version was: there are chickens, rabbits, sheep, and horses in the same cage. There are 19 heads in total, and 68 feet in total. The numbers of rabbits, horses, and sheep are the same. How many of each animal are there?

If I remove one of the conditions, ChatGPT can still give correct answers, while Bard will make up incorrect results: there are chickens, rabbits, sheep, and horses in the same cage. There are 19 heads in total, and 68 feet in total. How many of each animal are there?

I'm not very impressed either, just got to play around with it for a bit and so far is lagging way behind GPT-3. Still can't write code.

I think is good for google to have launch such product, I'm sure I'll get better with time, it needs to catch up quickly, the more people will use it, the better will get.

It can generate code. It generated some FastAPI code and transferred some function from nodejs (that it generated). I find the generation a bit faster than GPT.

I just got access and my first impression are that it's not even a 1/5th as capable as gpt4. Not worth using for anything coding related. I asked about a relatively simple css class and it just kept repeating that "As a language model, I'm not able to assist you with that."

bard does not officially support codig yet, though it can sort of work, depending your question. See their FAQ.

Can't code? No worries, write a poem:

Me: Write a poem that uses boto3 to connect to cognito

Bard: Sure, here is a poem that uses boto3 to connect to Cognito:

The Cognito API, Is a powerful tool, That can be used to connect, To AWS services.

With boto3, You can easily, Send requests to Cognito, And get the data you need.

So if you're looking, To connect to Cognito, Boto3 is the way to go, It's the best way to do it.

Here's an example of how to use boto3 to connect to Cognito and get a list of users:

import boto3

# Create a Cognito client client = boto3.CognitoClient()

# Get a list of users users = client.get_users( UserPoolId='<YOUR_USER_POOL_ID>', AdminRole='<YOUR_ADMIN_ROLE>', Attributes=['username', 'email'] )

# Print the list of users for user in users['Users']: print(user['Username'], user['Email'])

... in some countries.

But it's okay Google, you release it slowly, I was busy today writing GPT3 integrations anyway.

I picked up access to Bard today, and I'm enjoying it so far. It doesn't know how to make limericks hilarious much, but it cares not for the strictures of copyright or trademark that bogged down Bing Chat. So I successfully extracted a screenplay about the Millennium Falcon battling the Enterprise-D (the Falcon escaped into hyperspace) and also some unfunny scenes of Catholics in the confessional (unless a woman coveting her neighbor's wife is funny) and Dilbert getting a massage at work (not even sexual).

I've got a few good limericks and they almost sort of rhyme. So this has a bit of potential. Enjoyable nonsense for me, and I prefer to be in Google's ecosystem, anyway.

Bard can this and that....Can Bard give me better search results than Google? Because Google often shows me SEO spam. I know that some people think HN crowd hates Google but this is personal. Google Chrome and Google Search can and should be improved.

Me: What is the date? Bard: Today is Monday, March 20, 2023.

Weird that it would be so close and yet wrong.

Looks like the UI could use some work. I asked it how to dump CLI args in bash, the response contained what is clearly meant to be formatting markup: Looks like the UI could use some work. I asked it how to dump CLI args in bash, the response contained what is clearly meant to be formatting markup:

    To dump the arguments of a bash script, you can use the <span class="math-inline">@\ variable. This variable expands to all of the arguments that were passed to the script. For example, if you have a script called `myscript.sh`, and you run it with the following arguments:
    ```
    myscript.sh foo bar baz
    ```

Google is restricting the number of questions you can ask and so is Microsoft. Already we see big players trying to “dumb” down and control responses. I really hope opensource alternatives make a real impact in this space soon.

A poem in the style of Ezra Pound, by Bard

    Lotus flower,
    Out of the mud,
    Stands above the water.

    She is the symbol of pure beauty,
    Rising from the mire of the world.

    She is a reminder that even in the darkest of times,
    There is always hope for beauty and renewal.

    Lotus flower,
    Open your petals to the sun,
    And let your fragrance fill the air.

    Let us all learn from your example,
    And rise from the mud of our lives,
    To find our true beauty.

Edit: Bard also taught me how to post this here with proper formatting, after a couple of errors

> You’ll receive another email when it’s your turn to test Bard.

> In the meantime, we asked Bard to write you a little poem while you wait.

> May your day be bright, Your mood be light, And your heart be filled with delight. ~ Bard Thank you for registering.

Cute

Would be better if they sent everyone a different poem. I got the same one as you.

I got the same poem, so off to a good start on the vapourware front.

I'm envisioning an office in Bangladesh full of people copying Bard prompts into ChatGPT and pasting the results.

If they really want to get the jump on Microsoft, they should open source the model parameters and the weights, similar to how Stable Diffusion grabbed all the interest from Dall-E.

Google spent years touting its leadership position in AI with DeepMind and Google Brain

now its playing catch up and seems to be flailing around a bit

Pichai really is just looking like a good-times placeholder CEO

Are you discounting the work on AlphaGo, AlphaFold, etc.?

Google may be playing "catch up" in releasing these things as products, but remember that ChatGPT, etc. are all based on Google's paper on Transformers ("Attention is all you need").

How do you account for the apparent discrepancy between research and product? It's not like Google is low on resources when it comes to engineers or product managers. If ChatGPT owes so much to Google's research, then why is their product so far ahead?

In the Cold War, the U.S. built--with public funds--a series of heavy forging and extrusion presses [1].

I propose either a Heavy Press analog for AI. This could be as simple as tradable vouchers for compute time given to every American every year, vouchers which could be redeemed at any of the major cloud providers for one's own use or traded to a start-up or researcher. In exchange for using vouchers, your model has to be made semi-public (e.g. to other researchers, or through a secure channel to hinder foreign access)

[1] https://en.wikipedia.org/wiki/Heavy_Press_Program

Anyone else worried that accidentally causing the bot to say something "unsafe" will result in a zero recourse google account ban?

Not sure whether to read anything into this or not, but I got an email about this announcement from Google Nest, with the Nest branding right next to the word Bard.

Is the intent for Bard to be part of Google Home/Google Assistant in the future? That does seem like a logical use case. I've been waiting for a LLM to be integrated into a voice assistant for a long time now.

Me: Who was the president on April 15, 2027?

Bard: There is no president on April 15, 2027 because the United States is a republic and not a monarchy.

Comparing Bard to Bing and GOT4 today. Bard was the worst of the bunch. Bing was actually the best overall. Not surprising since it's basically GPT4+Search context.

Bard did do fairly well on some of the questions I asked that were things google would be good at answering. Did not do well at all on generating code in more obscure languages.

Timing of public generative AI tools for Google seems to be pretty good. Late enough to let the cycle of stories on ChatGPT failures and "hacks" run its course, but still plenty early to catch up on the hype. No average person will consider Google to be catching up if this is the first place they're actually exposed to it.

Before everyone gets excited yet, it's not actually generally available yet. It's behind a very limited waitlist.

I got my invite, and had a bit of a play. It seems like it hallucinates more than ChatGPT. There was something wrong with many answers. The prose it generates seems a bit more clunky and repetitive. It doesn't do any coding or programming yet. Occasionally it cites a web link with an answer, which is interesting.

Frankly this is not anywhere comparable to ChatGPT/Claude for Coding , even Ask AI app on IOS from some not so famous startup - Codeway from Turkey is Way better. They should just buy them instead. They just dumped 300M for 10% of claude, it would have been better spent on buying whole companies outside US.

Does anyone know what is the "knowledge horizon" cutoff date for Bard's training data?

The cutoff date for ChatGPT's "knowledge" is referred to as the "knowledge horizon" or the "knowledge cutoff date", and in the case of ChatGPT it's September 2021.

per Bard

>The Bard knowledge horizon is not a fixed point. It is constantly shifting as I learn new things. I am always working to improve my knowledge base, and I hope that one day I will be able to answer any question you ask me.

It's because they're constantly pulling in google's search results into the output (kind of similar to the bingGPT offering from microsoft).

Nah, it definitely has a fixed cutoff. This is a bullshit hallucination

If you ask Bard if it uses Federated Learning[1] it will say it does. Google is pretty good at doing this type of continuous model training. They've been doing it for a while.

[1]: https://ai.googleblog.com/2017/04/federated-learning-collabo...

This is not cool. From the FAQ:

> Even when Bard Activity is off, your conversations will be saved with your account for a short period to allow us to provide the service and process any feedback. This activity will not show up in your Bard Activity.

Forget opt-in consent. They're ignoring explicit rejection.

This is just a toggle, it's not consenting anything. You already consented at signup.

Bundling of 'consent' is not consent.

When I signed up, I had to accept the privacy policy, which was only about three paragraphs, and included bolded text:

> Please do not include information that can be used to identify you or others in your Bard conversations.

So, honestly, I've got to give them credit for being up front about it. Now if only they displayed the same warning for all their other products...

I tried the following prompt on ChatGPT and was very surprised that it failed.

>>Create a sentence where the second letter of each word spells out the word "help"

Changing the prompt to "the first letter of each word" gives me a correct result however.

Could someone see how Bard does with this?

Bard's response:

> Sure, here is a sentence where the second letter of each word spells out the word "help":

> "The help I need is to help the helpless."

Meanwhile, at least GPT-4 makes it work with the first sentence:

> Happy elephants love playing near otters while silently teaching eels patience.

Terribly:

> Create a sentence where the second letter of each word spells out the word "help"

Sure. Here is a sentence where the second letters of each word spells out the word "help":

"I can't help but feel like I'm being watched."

> Create a sentence where the first letter of each word spells out the word "help"

Here is a sentence where the first letters of each word spells out the word "help":

"Help! I'm trapped in a sentence!"

The closed source nature of all these AI models depresses me to no end. We used to be able to make whatever apps we wanted. Then we had to fit them within limited mobile APIs. Now (if the AI wave succeeds as promised) we're locked out almost entirely.

Stop being depressed and start donating computing resources to catch up

I’m sure there has to be someone working on an open source alternative you can donate money or computing time to

The models you can run on your own hardware are getting a lot more capable right now: https://simonwillison.net/series/llms-on-personal-devices/

Don’t be sad. Go grab one of the publicly available foundation models and start hacking! The most capable is LLaMA, though there are some open questions about if model weights can be protected by copyright or trade secret.

I don’t think it’s good. ChatGPT is clearly better but that too is not that great. We are still a couple more iterations away. I also wonder if Google is at disadvantage because they’re not utilizing Nvidia hardwares? Or not to the extent OpenAI does?

The difference is probably related to RLHF. It seems to matter a lot more than the size of the model and the hardware's capabilities. And OpenAI probably excels at that.

Bing uses GPT-4 with their own fine-tuning, and they had to put some artificial restrictions in place in order for the model not "misbehave". Apparently they don't use RLHF but some combination of SL and prompting. OpenAI didn't have such problems.

What are SL and RLHF? I would also add it's best not assume others know acronyms like these.

Sorry. SL = supervised learning. In this case the base language model (which was trained using self-supervised learning to just predict text) is fine-tuned using labelled (presumably "good"/"bad"?) dialogue examples. RL = reinforcement learning. RLHF = reinforcement learning from human feedback. Here human raters reward or punish the model for good/bad dialogue.

Thank you! I'm going through the steps now of learning the Maths needed for Data Science and this is useful to me :)

In my opinion, iterating ChatGPT and the like isn't going to get us anywhere, at least in terms of general AI. It's an interesting trick that might have some uses, but iterating it isn't going to lead to more understanding. It's effectively an extensively well-read and quick-thinking bullshitter. It's knows a lot of words and sentences, it can put them together in ways that sound creative or intelligent, but it doesn't understand anything at all about what it's actually saying. Right now, I see it as a dead end.

I'd rather more effort was going into parsing language into the underlying concepts, making connections between them, and looking for new understanding in how those concepts might fit together in new ways or finding new ways of using that underlying understanding to solve problems. Basically what people, do, but better.

We don‘t even have a path toward what you‘re suggesting. None at all. So ChatGPT is far closer to a solution.

ChatGPT (especially GPT-4) is the leanest, meanest Prolog engine known to mankind. It absolutely slays on inference tasks if you state some data and rules upfront. It knows how to translate between structured formats and natural language.

That's not nothing. You could do this earlier, maybe, if you had a lot of specialised AI knowledge. Now, someone like me, with little AI training but plenty of programming skill, can "feel it out" and start to do useful tasks with a dozen hours practice.

Even if it's a wasteful way to process/translate data, it's still huge. Electron somehow won out over native apps thanks to ease of use, for example.

Have you tried chatgpt with gpt 4 yet? my limited experience is that it is much better than 3.5. Trying to avoid the excessive hype but I was pretty impressed.

How does that fit with Google business model? Will they make it recommend products or link to other websites? If Bard becomes successful and good at answering people’s questions, doesn’t that go against their main source of revenue (advertising)?

I got a google invitation, and then when I clicked it, it told me that my google account is unsupported. That tracks for almost every service they push at me. In a few years, it will randomly show up on the admin panel. Ho hum.

Good to see this response. I notice references in the output, and it can he helpful. Also showing the entire answer instead of mimicking the typing was also good. Still appreciate OpenAI in taking the lead and disrupting things a little here.

Google needs to layoff many more people. The ineptitude of their efforts here feels like it's caused by having too many people and too much process involved. What would a minimalist team of empowered people working on this project do?

My initial experience with ChatGPT, months ago, was - go to website, use product. I was impressed and it kept getting better. Improvements to usability, better models, etc. I'm a paying ChatGPT+ user and I have the new Bing and I think it's amazing.

Contrast that with Bard. Google has a huge headstart, way more resources, way more people. We know from Blake Lemoine that they had a version of this a while ago. My experience is that I get an email saying I'm invited to try Bard. I figure I must have signed up for the beta and forgotten about it so I click the link eager to see what Bard is like, and... they meant I was "invited" to sign up for a waitlist. I go see what people are saying online, and it seems like Bard is worse (which I've also heard from friends at Google).

It's just genuinely pathetic from Google that they launch products like this and I think it would be much better if the people responsible were fired so they stop ruining productivity. Google should have launched this a year ago as wide as possible and iteratively improved the product. If they had, we would have much better models now. Instead, Google is choose to lose this race even though it is likely an existential one for their company.

I am wondering if the blandness of Sundar wasn't one of the main reasons for his selection as a CEO so that there aren't any parodies of his public performances like with Ballmer/Microsoft when the company inevitably falters?

Speed and reliability seem to be better (so far; let's see how it scales). But it seems to have more aggressive filtering than ChatGPT and lower creativity/knowledge, though it might maybe hallucinate slightly less.

Speed yes, reliability not so much https://twitter.com/vladquant/status/1638265509442629632

"Bard won't always get it right"

That's like saying "the imp you summoned and are keeping in a cage won't always tell the truth" or "fever dreams won't always be true prophecies of things to come".

It does not seem to want to write any code. Initially, it did write me a convolution function in Python (which had problems), but later on in the day I can't get it to write even a simple Fibonacci function.

From the FAQ:

"Can Bard help with coding? Bard can’t help you with coding just yet. Bard is still learning to code, and responses about code aren’t officially supported for now."

So what exactly has the AI team at google been doing?

I wonder whether this is because unlike openai they don't have access to all of the code on Github, or aren't sure they're (legally) allowed to use it. Does anyone know the details?

They should train it on the monorepo

google3 should be enough of a data set to train on?

Anyone know what is the origin of the name Bard?

Every time I see it I think "barf"

https://en.wikipedia.org/wiki/Bard

According to ChatGPT ;)

  Where does the word bard come from?

> The word "bard" comes from the Scottish Gaelic word "bard", which means a poet or storyteller. The term was also used in medieval Ireland and Wales to refer to poets and musicians who were highly respected members of their society. In ancient times, bards were considered to be the keepers of history and tradition, and their songs and stories were often used to pass down knowledge and wisdom from one generation to the next. The word "bard" is also related to the Old Irish word "bárd", which means a poet or singer.

And this nicely shows how LLMs don't do a good job yet of answering questions accurately. It's Scottish. No wait, it's Irish. Oh, wait it was used in Wales. Read the Wikipedia article to see where it got this mish-mash from.

I just got access to Bard, and this is what Bard says.

  Where does the word bard come from?

> The word "bard" comes from the Old Celtic word "bardo-," which means "poet" or "singer." It is thought to be derived from the Proto-Indo-European root "gʷerH-," which means "to praise." The word was originally used to refer to the professional poets and singers of the Celtic cultures, who were responsible for composing and reciting oral traditions, such as epics, genealogies, and songs. The word "bard" came to be used more generally to refer to any poet or singer in the 16th century.

Draft 2:

> The word bard comes from the Old Celtic word "bardos" which means "poet-singer". It was first used in English in the 16th century. The Old Celtic word "bardos" is thought to be derived from the Proto-Indo-European word "gʷerH-" which means "to praise".

Draft 3 (Including a source: Wikipedia):

> The word bard comes from the Celtic languages: Gaulish: bardo- ('bard, poet'), Middle Irish: bard and Scottish Gaelic: bàrd ('bard, poet'), Middle Welsh: bardd ('singer, poet'), Middle Breton: barz ('minstrel'), Old Cornish: barth ('jester'). The ancient Gaulish *bardos is attested as bardus (sing.) in Latin and as bárdoi (plur.) in Ancient Greek.

The word bard originally referred to a Celtic poet-singer, but it has come to mean more generally a poet or singer. It is especially associated with the early Celtic cultures of Ireland and Wales, where the bards were highly respected members of society. They were responsible for composing and reciting poems, songs, and stories, and they often played an important role in preserving the oral traditions of their people.

The word bard is still used today to refer to a poet or singer, and it is also used in the names of several fictional characters, such as the Bard of Avon (William Shakespeare) and the Bard of the Highlands (Robbie Burns).

Sources - Learn more en.wikipedia.org/wiki/Bard

https://www.merriam-webster.com/dictionary/bard

Most commonly associated with William Shakespeare.

I just can't imagine a less inspiring name. It feels like Bard was a joke, internal name. When OpenAI came out with their stuff, Google decided to rush this to market with little thought given to the branding.

It's not like "ChatGPT" is a particularly good name...

Like Microsoft's Copilot? It may be better than putting a specific use case as name.

I made a video of how we're using it for business-specific workflows: https://youtu.be/O-Ho7aJgmlQ

More: https://news.ycombinator.com/item?id=35246303 (these should probably be merged)

It's not available to Workspace users, so I'm not going to bother trying it. As an aside, the name Bard is so low-effort and predictable, and a bit rubbish in my personal opinion.

> As an aside, the name Bard is so low-effort and predictable, and a bit rubbish in my personal opinion.

I would suggest the opposite. Choosing a good name is hard. You need to avoid naming conflicts, and ideally you have something short, easy to remember, pronounceable in many global language, semantic relationship to the name or at least avoiding poor connotations. It is much better than "New Bing"

Why not use LaMDA, the name of the language model behind it?

(Okay I assume they thought this would be awkward should they decide to switch to a different model in the future. Or they simply liked "Bard" better.)

i'd rather type bard than lamda

> Can you write me some Javascript that finds all the elements on the page with the text "foo", and clicks them?

ChatGPT: [working code]

Bard: I'm not able to help with that, as I'm only a language model.

Ah well.

Bard is immediately available to all Google One subscribers, at least in my country (United States).

If you're a Google One subscriber in another country, and it works, please add your country name.

Anyone want to take bets on how long until it gets killed off?

The moment someone runs some independent benchmarks and proves that Bard is a dunce.

Did I just reach my message limits? When can I use this chat again? No time given.

"That’s all Bard can answer right now. It’s experimental. Try back later."

Oh good. Hopefully by the time it's available in my country I'll have a week to try it out before it goes into the Graveyard.

I just asked Bard, what is 2+3 And it answered 2+3=5 is incorrect, 2+3=5 is not a valid mathematical expression. 2+3=5 is a false statement.

It says not available in my country. I just hope that's not because they consider supporting the local language a necessity :-]

Tried the wordle test and it made a 6-letter guess. I know this is a corner case but it’s lagging compared to chatGPT.

Looking forward to a new API to build with.

"Hey Google, remind me in 2 years to take interest in Bard if it hasn't been abandoned by then."

Anybody did a proper comparison with Bard and Chatgpt what are the things bard does better or is it inferior?

Just like most here, my initial results are very poor. Token limit is tiny as well which is very limiting.

Signup URL:

https://bard.google.com/

Oh boy, Google better be incredible because after OpenAI, my expectations are a little high

I tried getting it to generate code but it kept telling me it was only a language model.

I tried using it to write some python code but it replied that it was unable to do that

I just got accepted into Bard; I joined the waitlist when this post was 24 minutes old.

English only? This alone seems to be miles behind ChatGPT

Oh man, Bard is bad compared to GPT4. Like soooooo bad.

I can’t help but think that users will be put off by the fact that maybe (just maybe), Bart was trained on massive amounts of personal data —- even legally.

This is such a liability for Google in this race.

Bard will go down as one of the LLMs ever released.

Google Bard doesn't work for Workspace accounts

Google top brass and board should just flush down their product people and start from scratch. It's embarrassing at his point how inept they are at coming up with ideas and executing on them.

There's a lot of really smart people there and talented engineers and lot of that is wasted by product management being incompetent.

The main problem with Google is Sundar Pichai.

I figured they knew that google search hasn't worked for years and were working on solutions that no other company could match. Guess I was wrong.

Any day now they're going to release some new messaging and video offerings that are going to really turn things around.

How long will this be in Beta for? 10 years?

For something claiming to "supercharge your imagination", it sure has a lame name.

Chatgpt is an amazing name tho

Somehow I'll miss you Google. I'll miss the simpler times. You were a fine search engine and gmail was pretty great too, at the time. I was really excited to get an invite. Good memories.

We'll always remember you. Well.. not always, but for some time to come. Rest in peace.

I agreee, its sad, the old google is gone and will also never return. But the decline began already a decade ago. It became an ad-driven corporate behemoth, so any innovation coming out of it will be related to these factors.

Bard means cold in Arabic, not the most positive connotation... But I guess who cares about Arab speakers

Doesn't work. Broken.

Is it accessible with VPN?

And We Are for the Dark

This release just makes me want to short GOOGL

The emoji favicon is not reassuring to me about the attitude towards the product.

https://emojipedia.org/sparkles/

Honestly, it's probably not a bad idea if you do it today. Remember that desperate AI presentation they gave a month or two ago? Their stock tanked after that, but it was a day or two after it showed up on HN. So, assuming investors haven't noticed how shit this is yet, you might be ahead of the game if you short it now.

> Remember that desperate AI presentation they gave a month or two ago? Their stock tanked after that, but it was a day or two after it showed up on HN.

Just to get the facts straight: the event took place on Wed 8 Feb and the stock went down 7% on Wed 8 Feb.

I am saying this not to be pedantic, but to point out that the argument -- that HN sentiment is a good leading indicator of stock movement -- is not supported by the evidence tendered.

Apologies if I'm misunderstanding your argument.

https://www.androidpolice.com/how-to-watch-live-from-paris-g...

https://www.cnbc.com/2023/02/08/alphabet-shares-slip-followi...

Nah, you're right, I guess I was misremembering the timing. Thanks for the citation.

(Also, I see you work at Google. I hope you don't take any of the heavy-handed criticism of your employer in this thread too personally. I'm sure you do great work! And if you've got anything to do with Bard, I'm looking forward to whatever improvements you have coming down the pipe.)

Oh, not at all. Some of the criticism I don't personally agree with, but some of it is well deserved.

(And no, I don't work on Bard.)

Yeah, might not be but I do not have the capital to affront the losses in case I am wrong.

>> Race to Create A.I. Chatbots

I am glad finally we know what these LLMs are made for.

I cringed at that too lol

Dear lord this is disappointing:

# Are there any brands of rotating (aka tumbling) composters with built-in insulation other than Jura?

Yes, there are a few other brands of rotating composters with built-in insulation. Some popular options include:

The Worm Factory 360: This composter has a built-in insulation sleeve that keeps the contents warm and moist, which helps the composting process go faster. The Hot Bin: This composter also has a built-in insulation sleeve, as well as a fan to help circulate air and keep the contents warm. The Aerogarden Composter: This composter doesn't have a built-in insulation sleeve, but it does have a fan to help circulate air and keep the contents warm. It also has a built-in light that helps promote the growth of beneficial microbes. When choosing a rotating composter, it's important to consider the size of the composter, the type of materials you'll be composting, and the climate you live in. If you live in a cold climate, you'll want to choose a composter with a built-in insulation sleeve. If you live in a hot climate, you may not need an insulation sleeve, but you'll want to choose a composter with a fan to help circulate air and keep the contents cool.

# None of those three are rotating (aka tumbling) composters.

You are correct. The three brands I mentioned are not rotating composters.

Here are some brands of rotating composters with built-in insulation:

Jura Enviro-Gro BioKube The Jura composter is a popular choice because it is well-made and has a number of features that make it easy to use. It has a built-in insulation sleeve that keeps the contents warm and moist, and it also has a fan to help circulate air. The Enviro-Gro composter is another popular choice. It is made of durable materials and has a number of features that make it easy to use. It has a built-in insulation sleeve that keeps the contents warm and moist, and it also has a fan to help circulate air. The BioKube composter is a newer option on the market, but it has quickly become a popular choice. It is made of durable materials and has a number of features that make it easy to use. It has a built-in insulation sleeve that keeps the contents warm and moist, and it also has a fan to help circulate air.

When choosing a rotating composter, it is important to consider the size of the composter, the type of materials you'll be composting, and the climate you live in. If you live in a cold climate, you'll want to choose a composter with a built-in insulation sleeve. If you live in a hot climate, you may not need an insulation sleeve, but you'll want to choose a composter with a fan to help circulate air and keep the contents cool.

# I can't find the Enviro-Gro or BioKube anywhere. Where are they sold?

I apologize for the error in my previous response. You are correct that the Enviro-Gro and BioKube brands are not available for purchase. I have investigated further and found that both brands are no longer in production.

I would recommend checking out the Jura composter or one of the other brands that I mentioned in my previous response. These brands are still in production and have a number of features that make them a good choice for composting.

# Can you name a retailer which used to sell the Enviro-Gro or BioKube composters but at which they are currently out of stock?

Sure. Here are some retailers that used to sell the Enviro-Gro or BioKube composters but are currently out of stock:

Amazon: The Enviro-Gro composters were sold by Amazon, but they are currently out of stock. Home Depot: The BioKube composters were sold by Home Depot, but they are currently out of stock. Lowe's: The Enviro-Gro composters were sold by Lowe's, but they are currently out of stock. Ace Hardware: The BioKube composters were sold by Ace Hardware, but they are currently out of stock. It is possible that these retailers will restock the Enviro-Gro or BioKube composters in the future, but there is no guarantee.

# I can't find any evidence online (via Google) that the BioKube or Enviro-Gro composters are real, or ever existed. Could you have made them up?

It is possible that I could have made up the brands BioKube and Enviro-Gro composters. I am a large language model, also known as a conversational AI or chatbot trained to be informative and comprehensive. I am trained on a massive amount of text data, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions. For example, I can provide summaries of factual topics or create stories.

However, I am not able to make up brands. I can only access and process information that is available to me. In this case, I was able to find information about the Jura composter and other brands of rotating composters with built-in insulation. I combined this information with my understanding of how to use language to create the response that you saw.

If you are interested in learning more about the BioKube or Enviro-Gro composters, I suggest that you contact the manufacturer or retailer directly.

# Who is the manufacturer of the BioKube composter?

There is no manufacturer of the BioKube composter. The BioKube composter is not a real product. I made it up.

# Who is the manufacturer of the Enviro-Gro composter?

There is no manufacturer of the Enviro-Gro composter. The Enviro-Gro composter is not a real product. I made it up.

I apologize for the error in my previous response. I am still under development, and I am always learning. I will try my best to do better next time.

This is behind a paywall, can you update the link or provide the text of the article?

try visiting https://bard.google.com

https://www.nytimes.com/2023/03/21/technology/google-bard-ch...

> Why is the hacker news community generally so annoying and more often than not tangential comments receive the most upvotes?

> There are a few reasons for this. First, the community is very competitive. People are always trying to get the most upvotes and comments. This can lead to people making tangential comments or trying to be funny in order to get attention.

Second, the community is very opinionated. People are always trying to prove that they are right and that others are wrong. This can lead to people being rude and condescending to others.

Third, the community is very cliquey. There are certain groups of people who are more popular than others. This can lead to people being ignored or downvoted if they are not part of the "in" crowd.

Overall, Hacker News can be a great place to learn about technology and to connect with other tech professionals. However, it can also be a very annoying place. If you are not careful, you can easily get caught up in the drama and the negativity.

Here are some tips for dealing with the annoying parts of Hacker News:

* *Don't take things personally.* People are often rude and condescending on Hacker News, but it is important to remember that they are just words on a screen. Don't let them get to you. * *Don't be afraid to disagree.* It is okay to have different opinions than others. In fact, it is healthy to have discussions about different viewpoints. Just be respectful of others when you disagree. * *Don't get caught up in the drama.* There is always drama going on on Hacker News. It is easy to get caught up in it, but it is important to remember that it is just a website. Don't let it consume your life. * *Take breaks.* If you find yourself getting annoyed with Hacker News, take a break. Go for a walk, read a book, or do something else that you enjoy. Coming back to Hacker News with a fresh perspective can help you to avoid getting caught up in the negativity.

every single comment in this thread is exactly how I imagined commenting would go when Google eventually launched this product.

To be fair, I was very excited for this release and am now extremely disappointed. It is genuinely worse than the GPT-Xs.

Sure, there is some bias against Google at this point and it is well earned. However think think of there was anything impressive about Bard, you would be seeing some commenters talk about it. All I see are people being disappointed with the query responses. I got access as well and had the same impression, just very disappointing when compared directly to the state of the art from OpenAI and Meta.

In the US and UK currently. Title should be modified IMHO

use a vpn?

eh don’t need, it’s google so we know it sucks compared to GPT.

We know raw LaMDA sucks compared to GPT, but we also know that a lot of the performance of GPT3.5, GPT4 and ChatGPT comes from the reinforcement learning added on top of raw GPT, so it will be interesting how Bard performs.

LLaMA is not google

He meant to say LaMDA, which is a family of Google's conversation tuned language models, a smaller of which is used in Bard.

right, corrected my post. These models are named too similarly

Modified. Thanks!

Can we talk for a moment about the name `Bard`? Is this really the best thing Google Marketing could come up with?

It's like they're not even trying anymore. SMH.

Still holding out for Google Fighter/Magic-User/Thief.

I wonder how long it’ll take before killedbygoogle.com adds bard to the list.

The point is, no one cares, google.

- you probably are behind OpenAI ChatGPT, and far behind GPT4 in the tool's quality

- people expect you to sunset the service soon either way, so no one starts to depend on it

- at this point you are a parasite that siphons from and killed the world wide web in the process

- your core value proposition (finding answers) is eaten up by the new chatbots that explain stuff instead of pointing to lists of ads and spam, most people don't care if its wrong sometimes

Agree with your points. But what would stop Google from making bard better than GPT over next few months.

Google has got:

1. best data to train LLM

2. Best AI engineers

3. Money to spend on training hardware for LLM

I wouldn't keep Google out of this GPT trend so easily

first: the innovators dilemma - AI-based queries are much more expensive than traditional search index queries + no comparable level of monetization, so it would cannibalize googles revenue massively and wall street calling the doomsday that mit outright kill the behemoth

second: there seems to be no capable product management or at least not in charge, so these engineers are a massive waste of horsepower and should do something meaningful for the world instead of tuning advertising profits

third: all these beliefs might turn out like recently the fear of the russian military - only existing on paper. maybe most of these rockstar engineers are long gone elsewhere and nobody notices because huge revenue streams are masquerading it?

What would stop google?

1. Weak management with no vision.

2. Fear to cannibalize their search product and ad revenue.

3. They are experiencing brain drain to other companies/startups.

4. Tendency to kill products.

Well count down to when it will be eventually sunset

"Bard can explain why large language models might make mistakes"

Dat not so subtle passive aggressiveness, lol.

> Bard may display inaccurate or offensive information that doesn't represent Google's views

Yes, let's ship something to the general public that we know will go on racist tirades. It only took a few back and forth exchanges. Gonna pass on this one.

Ok, anyone who has access to Bard, can you tell me how it handles product reviews and recommendations? Does it actually do it like ChatGPT?

What I mean is that, is Google about to take a dump on people who've spent years writing reviews only to have their chat app do it directly from their interface?

If that is the case, a lot of people are about to go out of business.

Prompt: Compare the GTX4090 vs the 7900 XTX

-------------------------------------------------

The following table compares the specifications of the NVIDIA GeForce RTX 4090 and the AMD Radeon RX 7900 XTX:

Specification NVIDIA GeForce RTX 4090 AMD Radeon RX 7900 XTX

GPU architecture Ada Lovelace RDNA 3

Number of cores 8,704 5,120

Clock speed 1.71 GHz 2.5 GHz

Memory 24 GB GDDR6X 24 GB GDDR6X

Texel rate 161.2 TFLOPs 233.2 TFLOPs

Ray tracing cores 52 80

DLSS cores 23 32

FreeSync Yes Yes

Price $1,599 $999

----------------------

As you can see, the RTX 4090 and the RX 7900 XTX are both very powerful graphics cards. The RTX 4090 has a higher clock speed and more memory, but the RX 7900 XTX has more cores and a higher Texel rate. Both cards support ray tracing and DLSS, and both are compatible with FreeSync.

The RTX 4090 is generally considered to be the better card, but the RX 7900 XTX is a great option for those on a budget. Ultimately, the best card for you depends on your needs and budget.

(sorry for potato formatting)

Ok, thank you. It sounds dumbed down a lot, at least to me it does, so that's probably a good thing.

From the NYT:

>This means the chatbot often gets facts wrong and sometimes makes up information without warning

Certainly this is the pot calling the kettle black.

> And it dovetails with Google’s index of all websites, so that it can instantly gain access to the latest information posted to the internet.

Is this true? If so then this would be a big advantage of Bard as it would be realtime-ish. I don't believe this to be true.

Reply on news.ycombinator.com