Even Apple cannot explain why we need AI in our lives
ft.com
ft.com
Code completion is a great use, some writing tools are nice, I love chatGPT for information lookup (while knowing that it hallucinates), ChatGPT/whatever is great for generating icons and some images, and it’s much better at writing in corpo than I am.
But…Nvidia as the most valuable company in the world? $1B valuations for freshly minted companies? Radical changes to the economy? AGI?
It certainly doesn’t seem that way. Yes incremental (in this case a large increment!) improvements in the way we interact with machines.
But humans stay winning.
They really don't need to walk back anything. If the features are too expensive to support and ultimately unused, then they'll just quietly kill them off. Otherwise, it'll just be another feature in the feature lists. Not something that really buys anything but also not something making them a lot of money.
As for crypto, I probably should have said blockchain instead. There was one point recently where everything was block chained.
Web 2.0 was "social web", wikis, AJAX requests, and the move towards SPAs. It's very much still the reigning paradigm.
- https://en.wikipedia.org/wiki/Web_2.0
I don't think I need to elaborate this quotation?
But if I do, note the domain I'm quoting from and the website in which you are reading this comment from me.
Big tech wasn't invested in crypto in the way it is in AI. For a few days Meta pretended they were going to launch a cryptocurrency, but that's kinda it. Whereas in this case all the biggest tech players are making a lot of noise about how AI is The Future and investing a tremendous amount of resources into trying to make their AI be The Future.
Big tech investment into machine learning didn't need walking back because machine learning morphed smoothly into AI.
Big tech investment into "Web 2.0" didn't need walking back because "Web 2.0", to whatever extent it was a definite specific thing, was a huge success and now we just call it the web. (If you meant "Web 3.0", that didn't need walking back for the same reason as crypto didn't: it was never a very big thing for very big tech.)
So I'm not seeing how any of those things was like AI today. If it turns out that present-day AI is a bubble and it bursts, it will not be like crypto or "Web 3.0" for the big tech companies because those were only ever fads that the big tech companies didn't particularly tie themselves to. It won't be like machine learning because it will be the same bubble, just later on. And it won't be like "Web 2.0" because that wasn't a bubble and didn't burst.
With blockchain technology there were many people prepared to die on that hill. It was going to solve world hunger albeit not the obesity epidemic. It also had all of its own lingo for not selling or buying the dip.
I remember people saying how things would be replaced with blockchain technology such as ecommerce, and apps 'on the chain' would find many thousands of uses. People were quite insistent.
On HN you risked getting modded down to hell and shadowbanned for life if you said anything wrongful about crypto. It was bizarre.
With AI we don't have this level of zealotry. You can say it is a load of rubbish and nobody wants to kill you. Weird, but indicative that AI has found some use, even if it is for 'art' and just marketing.
Nonetheless, if someone does get a bit zealous about it, I can tell them that I have all problems solved already with blockchain technology apps, just for trolling value.
Or they’ll use it to make stuff like Cortana (or whatever they are calling it now) slightly less useless.
The most vulnerable Big Tech organization are the ones like OpenAI and Anthropic that are swollen with investment specifically for AI but need to find a business or growth plan that can sustain that investment, plus the smaller and startup companies that have bet the farm on AI before it matured. But these same companies are the ones that have the potential to win big if there's another key breakthrough on either capabilities or product insight.
Because current AI is effectively being pitched as assistants 2.0, I don’t see them wiping AI from their lineup, but letting it rot is a possibility as they all have done for their v1 assistants.
Due to the attention economy, most aren’t going to care within a year. Google is still a dominant company despite killing off beloved services so if they can do that, the others certainly can weather failed AI products.
There might be some exceptions like Adobe, it depends on what they do. That’s my take on it anyways, in general there was just way too much pumped into AI without much thorough thought. The fact that Rabbit was even a thing people got hyped over speaks volumes to how out of touch with reality current AI pushes are.
It will be some time until we get to grips with all the implications of computers grokking language. Then again sci-fi has had computers understanding speech for ages as if it's the most natural thing in the world.
Hallucinates?
I'd use the term 'bullshits' when it doesn't know the right answer.
This makes it a dangerous thing to trust.
It is a greatat summarizing but poor at understanding. It is great at concepts but crap at precision. And every actionable "fact" coming from it needs to be vetted.
"Bullshits" implies it knows what it's doing, and can have a positive connotation depending on context.
The quality of the LLM matters. The better LLMs don't hallucinate as much as the rest. ChatGPT-4 appears to be trained to consult the web when it receives questions for which it's likely to hallucinate, such as questions about hard figures.
Eh? Only a sociopath, when confronted with a question they don't know the answer to, confidently makes something up. The human may not know the answer, but they generally know whether they know the answer or not. The machine does not (of course, it doesn't _know_ anything). You should absolutely trust something less than this far less than the average human.
What? No. Literally fake memories are such a well-known phenomenon the cops will abuse it to generate fake solutions to crimes etc.
https://en.wikipedia.org/wiki/Confabulation
Also there are people for which compulsive lying is a pathological problem, who are definitely not sociopaths, and there are many people who demonstrate the processes to non-pathological/non-clinical levels. People will, indeed, “just make it up” perhaps especially if they don’t really know the answer, because that’s socially embarrassing etc - that’s literally the opposite of sociopathy.
https://www.goodtherapy.org/blog/psychpedia/compulsive-lying
5 years ago, just getting a computer to form vaguely relevant grammatically correct sentences felt magical.
A lot of smart people _always_ seem to think <insert tech> won't fizzle out.
Please. My dad voted with his wallet for… Video 2000.
(He also advised me to buy Lloyds shares when they were about x10 their current price).
I really don't think crypto or VR or AR are going to fizzle, I think they're just early in their overall lifecycle. I absolutely think that in ten or twenty years those categories will be much larger than today.
I guess maybe VR is another one? But that's another example of a slowly growing ecosystem.
I think mostly the people who think some piece of hot tech is going to take over the world aren't usually wrong, they're just early.
Ruby on Rails. Modula-2. Every JavaScript framework since ever. Supersonic passenger flight. Space tourism. Emitter-coupled logic. Gallium arsenide (still has my vote for a name for a speed-metal band). Musk's Boring Company. I'm sure I could go on with some more thought.
But regardless, I don't think that many people really thought Musk was going to build much with Boring. I certainly didn't.
And has anyone in the last three decades (other than Boom's hapless investors) thought that supersonic passenger flight was going to be huge?
But it's also why Lord Kelvin claimed in 1895 "heavier-than-air flying machines are impossible"* and in 1897 that "radio has no future".
* don't ask me why birds were not a proof by example; we have the same problem today with regards to human brains being proof by example that human level intelligence is possible.
It seems like one of the defining characteristics of a paradigm-shifting technology is that a lot of smart people dismiss it at first. The skepticism and cynicism I see about AI from the HN crowd (of all places) is such a great example. I use AI every day as both a user and developer and it's an incredible gamechanger. It's pretty surprising to me how incurious and conservative many HN commentators appear to be, but maybe it shouldn't. I'm guessing the median age on this site of the active commenters is now in the mid 30s or early 40s.
https://en.wikipedia.org/wiki/ELIZA dates from 1964.
Objectively, no it doesn't. Even if you think it's a bubble, then the bubble is just growing. There are zero signs of anything popping yet. Company valuations are only increasing.
> Nvidia as the most valuable company in the world?
It makes sense when you look at the financials. It's like back in 1800's when railroads were the new thing, and imagine there were only one company that knew how to make locomotives, while everything else had multiple suppliers. It would have been the most valuable company as well.
Nvidia is in an extremely enviable situation where they're simply the only company that makes a product that is in extreme demand. And for various reasons, no other companies appear to be able to make and deliver a substitute -- not yet and seemingly not in the near future.
So they're able to produce an extreme level of profit, which shows no signs of stopping anytime soon, and their market valuation reflects that accurately.
Remember, high market valuation isn't a reflection of expected future revenue or sales -- it's a reflection of expected future profit. And Nvidia is raking in crazy profits because nobody else is effectively competing with them.
AMD ""just"" need to fix their software API ecosystem. Developers, developers, developers. Unfortunately I don't think they can do that even if promised a trillion dollar stock valuation. Not because it's technically difficult, either.
Yes. NVIDIA's customers are mostly the cloud providers buying tens of thousands of machines each. The latest 8xH100 pods cost almost $400,000 so no one but Facebook and a few other tech giants can afford to buy tons of them without having customers lined up and ready to rent them.
If you look at AWS's reserve pricing, a one year contract costs more than the hardware would if you bought it up front. The on-demand pricing is even crazier, if you're lucky enough to find some capacity. NVIDIA's customers are absolutely printing money, it's their customers' customers that are paying for it.
All that matters is that people think the industry will turn out to be profitable and sustainable overall, once we figure out which products and features work and which don't. There's no reason whatsoever to think AI doesn't fall in that category.
NVIDIA, specifically, doesn’t really make a compelling product, but is trying to make it seem like their tools can eventually make a compelling product, so build your platform on NVIDIA so that you can sell the opportunity that allows compelling products…
Ever seen a bubble pop? It doesn't contract and then pop, usually. A hole develops, and almost immediately no more bubble. If you watch very closely, you might see signs of the end a fraction of a second before it vanishes.
Like, this is why these phenomena are called bubbles; they grow and grow and grow, and then bang.
> It's like back in 1800's when railroads were the new thing, and imagine there were only one company that knew how to make locomotives, while everything else had multiple suppliers.
I, er, feel obliged to mention that that was a particularly notorious bubble. And, like, yes, if you're selling shovels in a gold rush, that's a good place to be, no question of that. But it doesn't say anything in particular about the long-term viability of the gold rush.
There's no "it feels like it's popping".
The bubble is just continuing to grow. When it pops, you'll know it.
(Although to be fair, it's not instant. It's often a jagged decline that takes weeks or even months. You never know when you've hit the bottom, when it's "fully popped". So the metaphor with actual physical bubbles is not perfect. But in hindsight you can often look back to a day of massive selling that "started" it.)
> Ever seen a bubble pop? It doesn't contract and then pop, usually. A hole develops, and almost immediately no more bubble. If you watch very closely, you might see signs of the end a fraction of a second before it vanishes.
Well, if that’s the case then why is the OP arguing that the bubble is popping based on feelings and vibes instead of objective data caused by the rapid popping of the bubble that you postulate?
Literally you’re saying the bubble is going to pop so fast that it’s noticeable to everyone and objectively undeniable in fact, the numbers are going to show it, and that’s why… grandparent’s vibes are evidence that it’s happening, quickly and indisputably, despite the failure of the numbers to show it?
Incoherent take.
I wonder how much of this has to do with 15+ years of the narrative of AGI/"The Singularity" being near. Deep learning seemed to really reinforce the narrative as it has blown away classical machine learning and succeeded in so many applications. It started to look like all of Kurzweil, et al's predictions were dead on. Particularly as LLMs took off.
Of course, we're now seeing the apparent limitations in very difficult applications like reasoning as LLMs have mostly been revealed to be a dead end path towards reasoning (with apologies to Blake Lemoine ;) ). And self driving is being revealed to be extremely, extremely difficult to solve with pattern matching alone.
Even the traditionally conservative Apple seems to have bought into the hype (though some of the applications will be pretty useful, IMO, after a few generations of release).
I saw somewhere (X I think) someone made the case that the latest generation of AI buzz has only created between $5-$50 billion in value despite investments measured as a multiple of that. I found it convincing.
Personally, I'm still a believer that AGI will happen, but I am convinced by Yann Lecun's take. We're not even on the on ramp yet, and all the bluster and hand wringing over safety threatens to kill or tragically stall the most important technology in all of human history. [1][2][3][4]
[1] https://x.com/ylecun/status/1662020967529435140
[2] https://x.com/ylecun/status/1801018194121118103
Yes, there's a bubble.
Those things you list, I find it very easy to believe they add up to at least 3% productivity boost globally. A single tech doing this in one year (or even 18 months) before we really understand how to best apply it, is a radical change all by itself, even if we don't use it to speed up future growth.
The world economy is $109 trillion; 3% of that is thus just over $3 trillion per year, which is enough to justify NVIDIA's valuation… assuming they get to collect all the profits and nobody else manages any, which is unlikely given their hardware burns a lot of expensive (in aggregate) electricity.
I can also easily believe that this triggers a painful demand increase for electricity.
I hope it isn't economically dominant, because nobody is ready for that transition.
I find it incredibly hard to believe that these things provide value anywhere close to 3% of GDP, that estimate is orders of magnitude too high.
3% is the entire economic output of the UK or India. The global agriculture industry is only like 4.5%.
I can't find out how many copy editors and copy writers there are in the world (I keep finding numbers for the USA alone).
I don't know how many graphical artists there are, nor how many are willing to use the various image generator AI, but I am seeing that stuff pop up IRL.
We know at least some lawyers have used it because they didn't check the output for hallucinations and it bit them hard; we don't know how many use it carefully.
>Nvidia as the most valuable company in the world?
This could just as easily be written as:
Apple as the most valuable company in the world?
Microsoft as the most valuable company in the world?
Tesla as the most valuable company in the world?
I won't be the one to defend Nvidia, but I don't find any solid arguments to consider its business model worse or better than those of the 'other' IT companies.
But they are an interface between the real world and computing, because everything in the real world revolves around human language. Here are a few things that are actually solved by now, even if we did not yet fully recognize it:
1. automatic human-human translations
2. collection and annotation of all kinds of real-world data (with the help of image recognition, of course)
3. human-machine communication (not decision-making, mind you.)
The most crucial aspect is that other human-machine interfaces will become less relevant. In particular web pages and mobile apps. They might stay as an attempt to capture attention, but they will become less and less relevant for actual usage. Here are a few services that might get by without a web browser or app soon:
1. online banking
2. delivery services (and others like Uber)
3. small business that basically only need to manage appointments online (doctors, lawyers, car shops)
> 1. online banking
??! You've got a system of unpredictable reliability and you're going to let it handle money?
Online banking has always been extremely conservative, sometimes for good reason.
Yes. Recall phone banking? I imagine it like that. Potentially via a chat service like Whatsapp or signal.
You'd start with simple standard tasks and low limits, if course. But the risk and value can be directly quantified in online banking.
Banks indeed look at things in terms of risk mitigation, but this is silly. You will not see a credible bank incorporate LLM-based features because it is a genuinely enormous attack surface with marginal ROI. Any logical developer would create an ELIZA-style on-rails system instead, and it would never misbehave like ChatGPT.
You can't make money fulfilling 'needs' unless there is also demand, but you can certainly make money fulfilling demand regardless of what some newspaper decides you 'need'.
(I would imagine the next FT headline will be 'Starbucks can't explain why people need coffee', after all why spend $4 and wait in line when you can just take $0.01 caffeine pill.)
I know that a "power user" can get much more out of these tools than me, but I still can't imagine what people would use that for, especially on a phone. People use their phones taking calls, watching TikTok, sending messages, not getting answers. "Turn on the light" thing works well enough in the old assistant. Which makes me wonder whether all those things Apple advertised make much different in that ecosystem.
(btw it seems that the Samsung/Google AI stuff never got as much coverage as Apple's thing?)
Any tactic the so-called "tech" company can use to get people to buy new computers. When the old ones are still working fine.
https://cc.bingj.com/cache.aspx?d=732027381763&w=c_PUW-0p05q...
For example I made a bookmarklet that summarizes articles linked from HN. So the way I have skimmed this article is by reading this:
Apple's new AI offering, Apple Intelligence,
seems unoriginal and derivative compared to
competitors.
Apple is partnering with OpenAI to integrate
ChatGPT, but the integration appears
halfhearted.
The article questions if Apple's approach can
truly unlock the potential of generative AI.
Make up to three short bullet points about what
the following page is about. Under no circumstances
make any of the bullet points longer than 40 words.
The page:
It is a very crude approach. Not sure how well it really covers the most important points. But I already like these summaries to decide if I dive deeper. And over time the summaries will only get better.
I think AI will be part of most UIs in a few years.
There are some small ways can enhance your life but consider this, we already had type aheads, document templates and generators. Adding an heir of randomness into these things might make them flashier but not necessarily more useful.
When OpenAI starts announcing products from all the data people dumped into their models, no one will be laughing. Enjoy the sherlocking while it happens to your company.
The ability to easily (and privately) double-check emails for typos and grammar mistakes before sending them out will genuinely be helpful (who hasn't sent a late-night message with a typo?). And if the quality of the text summaries is high, easy summary creation will also be useful for many people.
Once it becomes capable enough to act as an agent and starts asking you questions (unprompted by you) to confirm its actions, it'll slowly become a requirement to participate in society and remain productive enough to be taken seriously, like smartphones are today.
Most of my family uses iPhones and it's painful to watch them struggle with basic tasks. My mom tries to find out how to access gmail instead of me.com. My brother downloads yet another app to add metadata to pictures. No one at work seems to know how to upload pictures as JPEGs (I think they default to AVIFs now. Not a problem, but it was a super-secret change to the defaults, and easy to support on the server, but we didn't hear about it until our users started complaining.
I know there are people who positively love their iPhone and the status it apparently confers in their friends group. But I've never understood why a device from a secretive developer-hostile company is supposed to be some amazing tool.
It's a miniature computer with built-in wireless network and a poorly integrated phone app. It's innovation seemed to me to be how well they were able to convince Verizon to finance the phone so users wouldn't have sticker-shock.
So "apple intelligence" is just repackaging other people's stuff in a way that their users can digest? That's not a surprise. I'm just waiting for my family to start yammering on about how Steve Jobs just added AI to their phones and no one else has such features even though Steve Jobs is dead and LLMs have been a thing for YEARS.
HN really underestimates how important this is to billions of consumers.
Having said that, last year someone I knew went from following every Apple event to buying a Nothing phone (Android), because Apple were offering nothing new for unreasonable prices. The iPhone diminishing returns era is here. Even the US might fix the "wrong colour in chat" problem.
You can't neglect the fact that, despite their lack of innovation while sitting on a huge pile of money, Apple has brought a few good things in the last decade. For example, their use of ARM chips in Macs has influenced other companies to adopt this architecture, which in some ways is better than x86. Another product market they opened up with their "innovation" is the AirPods, which are, in my opinion, better for daily tasks than bulky headphones as they are easier to carry around and more discreet.
>But I've never understood why a device from a secretive, developer-hostile company is supposed to be some amazing tool.
An iPhone may lack some basic features, but I find the overall experience more polished than Android.
>I know there are people who positively love their iPhone and the status it apparently confers in their friends group.
My 2021 iPhone SE doesn't confer me any status, but I agree that some people just buy iPhones because they are expensive.
Anyway, in the end, it's about everyone's opinion. But Apple’s AI seems like overkill. I definitely don't want AI spilled into every app that is on my phone. I'm happy using the phone app, taking photos whenever I feel like it, and keeping in touch with my friends over WhatsApp/Discord/Messages.
Aged like milk...
There are serious discussions about multi-100B USD compute clusters and building multi-GW power generation. If we could replace all artists plus half of mathematicians and scientists tomorrow, it still wouldn't be worth the investment. Nothing short of summoning the Computer God will justify that.
Ie we can’t even use AI to come up with a new and creative tag line. We will reuse one from 20 years ago
That’s how cool we are
Soon people might figure out you technically need nothing except a lean-to for cover from the elements and some basic food.
Send signals between machines is a lot bigger and different than dump everything into a statistical model.
Even immediately after the .com bubble burst it was still obvious the web had a ton of potential for communication, knowledge sharing, etc.
The only thing AI seems good at is pumping out low quality "content", and there was already too much of that before.
We have a hardware company pretending that just adding more TOPS is going to get us to AGI.
We have AGI benchmarks that don't even test for AGI (how can you be 50% AGI...)
At some point it has to deliver insane value, and that point is soon because the numbers aren't adding up. People are realizing it's good at like 30 of the 3000 things its proposed to do.
While part of the 2023 hype cycle was built on doomer fantasies of a massive and proximate labor disruption, I don't see a lot of people in the news industry feeling like their "business model is threatened" by Recall and Apple Intelligence, or even by ChatGPT and Copilot. I don't think that's what's driving the coverage.
"Cheating at homework." Is a calculator cheating? Was using encyclopedias cheating? What about wikipedia? What about Wolfram Alpha? This mentality that one must do all work originally or it's cheating is not how learning works / should work. We should teach kids how to leverage all the tools they can to find the right answer and use critical thinking to be confident the answer is right. If we discourage learning to use these tools we'll actually make kids more likely to just trust the tools at their word when they do go sneakily use them.
"... and their jobs." I'm a software engineer and AI has made the process of debugging issues and implementing common patterns so much faster and less error prone. It's because I'm a highly skilled engineer who knows what to look for (same as when we browse stack overflow) and how to identify good and bad patterns.
It's not laziness, and the work must be high quality, and now I have a tool which helps me get more done in less time without sacrificing quality.
Pedagogy is not about marketing spin of cheating as "tools at their disposal".
There are often reasons for having learning exercises, and reasons when introducing tools to that.
When you go to the gym, you don't use a forklift for the free weights, and you don't slap a motor on the stationary bike.
Yes if the point of the exercise is to improve your mental arithmetic skills. It's a bit arbitrary but it's not complicated. This is the same way that steroids is cheating - the competition is defined such that PED are considered cheating, so using them is cheating.
Yes, if you are supposed to be learning arithmetic facts.
> Was using encyclopedias cheating?
Yes, if you are supposed to be learning how to find, evaluate, and cite original sources.
> What about wikipedia?
Often not accepted as a source in school classwork.
> What about Wolfram Alpha?
Same, all these things are great assistants for quickly performing tasks you know how to do, but they are a hinderance to learning if you don't know. And if you are using them blindly you won't know when they are wrong.
Do you think a student learns anything by copy/pasting a school report from Wikipedia? If you were starting out today, you would become a highly skilled engineer if you relied on CoPilot? Or did you become a highly skilled engineer by devoting a lot of time, thought, research, and experimenting to your craft?
https://9to5mac.com/2024/06/18/apple-vision-cheaper-halts-wo...
"Report: Apple halts work on Vision Pro, aims to release cheaper Vision headset next year"
Not to mention the reveal was extremely strange and almost felt like they were not releasing anything at all for the first 15 minutes before they showed it in action, almost felt very self conscious.
I feel that level of tagline should have only been invoked when they truly manage to ship something on even the baseline of their competitors today in a (wishfully) local and secure way.
Hence the WwDC was no live demos and all “future promised” features we have seen everywhere else already
The fact that there is a prompt asking for permission doesn't change that. Maybe some parts of the AI in iPhone don't use ChatGPT but others do, as advertised.
> Apple Intelligence doesn't use OpenAI at all. Siri and writing tools can tap into ChatGPT to compose text, e.g. create a story but that requires approval from the user.
https://www.apple.com/newsroom/2024/06/introducing-apple-int...
"chatgpt" has 26 mentions here. And the title of the relevant part is:
"ChatGPT Gets Integrated Across Apple Platforms"
"Apple Intelligence" is a strange marketing driven name because it implies to the tech naive that all AI in an iPhone is offline.
When in fact "Apple Intelligence" is just the name for some of the AI functionalities in Apple devices. Some functionalities will still send data to ChatGPT and perhaps other third party providers in the future.
But Apple gets to advertise "Apple Intelligence" as privacy conscious, offline, personal AI and still be technically correct. Classic.
Anyway, ChatGPT was a very small portion of the Apple Intelligence demo. The significant majority of the functionality they showed (all the non-ChatGPT stuff) was on-device or in Apple's Private Cloud Compute, the privacy story for which is quite good.
Apple's stock would have dipped dramatically if they held WWDC this year with zero mention of any AI products. They're duty isn't to us as consumers, it's to shareholders to protect and grow their investment. They just have to play the same game as the rest of the industry. They don't have to explain why consumers need it, that's not the point. The point is that they can't look like they're falling behind.