Google shows off what ChatGPT would be like in Gmail and Google Docs | Hacker News Reader

Google shows off what ChatGPT would be like in Gmail and Google Docs

arstechnica.com

93 pointsgixo63 comments

I'm looking forward to the day when the "help me write" and the "summarize" features are the inverse of each other.

You're sending an email. You enter "I'm on it" into the help me write box, and the AI generates a complete corporate-speak email using all the latest buzzwords.

Then the recipients get the email and the summarize feature shows them: "I'm on it".

Slowly, no one even realized, our jobs were replaced by AI talking to itself.

... which illustrates the pointlessness of a whole lot of business activities.

"I’m Going To Need Those TPS Reports ASAP."

Bullshit jobs get replaced by bullshit generators.

I had a "OMG, we live in the future!" moment reading this headline, isn't this what computers in sci-fi movies do (ok admittedly there aren't a lot of sci-fi movies where office work is the plot): you ask it to do something like yo would an assistant, and it delivers you the results. Something like "Computer, scan planet for lifeforms" on Star Trek.

Maybe someone should build a KITT from Knight Rider using ChatGPT (just for his sass, leave out the self-driving part, that's Elon's department (/s)).

Knight Rider style: https://arstechnica.com/information-technology/2023/03/gm-pl...

KITT talking scene is so relevant to 2023

Same. More generally, I'd like a model that infers the most likely prompt(s) from transformer-generated words or images. I would much rather know what the prompter was trying to do than what his model generated as a result, and treat whatever was generated as a lossy wire encoding.

Yeah, this has been a meme in the ancient times (2 months ago or so)

https://i.redd.it/wc8ne97ainda1.jpg

It's almost as verbose as XML.

LOL

XML is not verbose if you stick to using primarily attributes instead of elements. Unfortunately, that pattern never had popularity.

Attributes don't map well to standard collections, so it is clumsy to parse and use in most programming languages.

If the revealed preference of everyone is short, terse speech, why isn't business writing optimizing for that?

Because it’s not polite.

Just waiting for google to have a superior overall AI, but so phenomenally fucking up the product launch that it just fades into obscurity while getting eclipsed by its competitors.

Really I can't think of anything more google than that. No idea why shareholders still tolerate their crap.

we don't know if they have superior overall AI..

More Google than that: the product launch is wildly successful and they get a lot of traction, only to succumb to institutional ADHD and sunset it to free up resources to launch yet another chat app.

Honestly, I’d love to know what childhood trauma afflicted Google to be the way it is.

It's all that hiring of the best of the best.

You probably have a point. The best of the best penalty get bored quickly and want to build the new shiny thing. We need more average people that can take care of maintaining existing applications.

While I like to joke about Google products failing as much as the next guy, you've got to give them credit for mostly being focused on research and publishing their results publicly.

Other than, you know, a company that calls themselves "Open"AI, who now even consider it too dangerous (due to competition), to even release the number of parameters of their most recent model, much less talk about the actual research.

Why should we give them credit for failing to build a working product and instead spamming research papers to conferences?

Hm, and what do you think begat all of this LLM fun in the first place?

Very coincidentally I am trying to understand AI better, and I am reading "Regular Expression Puzzles and AI Coding Assistants" and it says on its first few pages:

Due to transformer deep neural networks mostly pushed forward to a 2017 paper called "Attention is all you need". Prompted by your question I searched where the people who wrote this paper work at, and... (unsurprisingly drum roll), a team from Google.

It is well-known that Google had world-class researchers working on AI for a long time now, and that they pioneered a lot of techniques used in the recent AI craze.

And yet they have nothing to show for it other than some words on paper. Meanwhile OpenAI is eating their lunch by offering an actual product that people can use and interact with. Nobody cares that Google did the research they used to make it. Talk is cheap.

Their AI efforts go towards google search and ads.

I don't think it's that well known, and for an advertising company, Google's marketing department doesn't seem very good. For instance, Google translate has been an ML model for more than a decade. They dropped a bunch of hand written code and replaced it with a model. The Google home assistants use an ML model to do their speech to text transcription. Google's tensor hardware is world class for ML research. But all people say is that Google search sucks now and wah they killed off Reader/Stadia/etc. Transformers will possibly have a bigger effect on society than Google search did and I'm doubtful Google will get proper recognition for that outside of circles like ours.

> Transformers will possibly have a bigger effect on society than Google search did and I'm doubtful Google will get proper recognition for that outside of circles like ours.

Do you think the general public gives Xerox PARC a lot of credit for modern computing? I don't think so, and it's fine. It will just be well known among technology enthusiasts and historians of technology.

It actually does bother me when people refer to Xerox as a failed photocopy machine company without acknowledging PARC, but I'm aware I'm in the minority here.

I also think The Dream Machine should be required reading in a history class in all CS curriculum, but I'm not in academia.

https://www.goodreads.com/en/book/show/722412

> had

It seems clear that the recent golden age of AI research is behind us. OpenAI fired the first shot, and now every company is scrambling to put every dollar out of their budget into productization/weaponization of the tech. There is zero incentive for anyone to publish papers, models and weights now that guarding them as trade secrets is infinitely more valuable. It isn't researchers calling the shots now but but PMs.

This guy gets it.

It will be a real name policy again. The AI will refuse to use any pseudonym you give it and it will insist on always calling your real full name.

It will be extra creepy because, you wont even be logged in, it will just identify you by using all the data it has about you.

Occasionally when you are chatting to it, it will blurt out some some embarrassing or private information about you.

I don't think it is fair to compare an MBA/Suit/Executive CEO type with risk taking Entrepreneurial/Founder CEO types. They have a completely different mindset and motives.

A suit essentially plays a defensive game. He can't take risks. CEO is just a job. The goal is to not ... things up and basically collect a hefty paycheck every year. You become a suit by being highly risk averse and if you take risks and if it doesn't pan out you will be kicked out.

A founder CEO type is essentially a wild high energy high risk charismatic type. You can't compare someone like SAMA or Zuck or Steve Jobs or Elon who won't take no for an answer and go all in to make their dreams come true.

Did you just compare Sam Altman to Zackerberg, Jobs and Elon?

> No idea why shareholders still tolerate their crap.

Shareholders have no option. Larry + Sergey are still majority shareholders iirc.

> Just like how Google put social features into every product back in the G+ days, the plan going forward is to build ChatGPT-style generative text into every Google product.

Ah yes, the famous G+ integration that has been widely acclaimed as a successful, long-lasting and profitable move.

You know what still pisses me off? It’s tiny compared to most things, but it was the first thing that turned me against Google.

The removal of the ‘+’ search operator.

Before Google+ was a thing, a leading ‘+’ before a search term meant “this must be in the results”. Then they hijacked it for G+ and made me start "quoting" "every" "important" "term" in my query.

Or maybe it was the Reader shutdown. Those both happened around the same time, right?

I miss it too, but it was the sharp end of the wedge in the dumbing-down of search query interpretation. Now they'll take every word of your query, find all the synonyms, word-stem everything (think "kicked" -> "kick"), and search for that instead of what you typed.

I know, some (the majority of) users don't know or care about the difference, and I'm glad that you can for now still enclose the things you mean verbatim in quotation marks, but I wouldn't be surprised to see that disappear.

The + came after that, to cancel that out.

> You know what still pisses me off?

Yes, because ChatGPT knows that is the most probable reply to the prompt.

Must be devastating for the marketing team at Google to have their new product Kleenex'd like this.

Lol I had the same thought reading that headline. I wonder if the author even realizes that not all generative AI is ChatGPT (probably not because the bar for 'tech journalist' seems to be quite low).

The author pretty clearly knows that not all generative AI is ChatGPT if you read even the first line of the article.

However, the vast non tech readership wouldn’t know what generative AI is, but nearly everyone and their mother knows what ChatGPT is.

So the headline is extremely effective in conveying what Google’s demo does while being understandable by the largest possible audience. It’s an extremely valid and appropriate headline.

I bet the person who wrote the article and the person who wrote the title are two different people.

Both did the jobs they’re supposed to be doing. The article writer knew what they were talking about, and the headline writer knew how to get clicks.

That's true a lot of places, but Ars writers actually write their own headlines. In fact they're required to write two, and the CMS will A/B test them for an hour before picking a winner. You can tell which one is the "A hed" because it's in the URL. This one is the A hed, so the author probably preferred this over the other one. (Not sure what it was.)

What's "Kleenex'd"?

Nevermind, I went and googled it.

First ChatGPT request - "Why can't my Google Workspace account do everything my @gmail account can?"

Stop it! He’s already dead!!!

My mind is blown right now. I have noticed differences between the two, but I thought they were just visual/UI differences. Is there somewhere I could learn more about how Google Workspace does not have parity with a free Gmail account?

> With Google Workspace Individual you get additional premium features including longer group calls and intelligent noise cancellation with Google Meet, and easy appointment booking through your own professional booking page.

from https://workspace.google.com/individual/