Amazon Is Gutting Its AI Division After Sustained Failure
j / k
|
Down or up a comment |
h / l
|
Previous or next link |
?
|
This help |
When I was at AWS, we had data access locked down very tightly. I was trying to train a model to predict how long various dynamic workloads running on behalf of customers would take. Just getting access to anonymized production data from a DB that my team owned took well over a month and had me compromise heavily on what attributes I could access. I also spent a decent amount of time in meetings with lawyers and compliance folks to explain and justify all our data classifications for PII, retention, access controls, etc. We would review our data policies quite frequently. We needed break-glass procedures to access prod data in our own systems, and it was uncommon to need to resort to that. Even when you broke glass, certain classes of sensitive data would still not be visible to you. On the plus side, We did get quite good at operating and maintaining systems without needing to see sensitive info.
If you contrast that with Meta's disregard for privacy or the way frontier labs disregard the law, it's easy to see why Amazon would struggle in this space.
[1] https://en.wikipedia.org/wiki/Annapurna_Labs
[2] https://en.wikipedia.org/wiki/Amazon_Lab126
From my interactions with them, is that they really care about customer success, and everything including AI is a means to an end to them.
I think building a cool product with undefined reliability and capabilities, then selling fantasies about what it will possibly be able to do down the line isn't really what they do - or what their customers are paying for.
Don’t know to comment about the other areas of their business but AWS has always been excellent for me.
A relevant metaphor is the nuclear arms race. Nobody wants a world full of nukes pointed at each other, but if the other guy is gonna do it...
The argument in this case is, we can't slow down, because China certainly won't, and do you want those guys to build the machine God first? So it's "rational" to maximize p(accidental_skynet).
I can suggest a prompt that will report how many founders part of Y-Combinator, who came from nothing, some immigrants from communist countries, have become ridiculously successful.
Have you tried to start a business in Pakisthan, Russia, Cuba or China as an outsider? Even with a few thousand dollars as seed? Do you want to buy a house in Pakisthan, Russia, Cuba? Do you want to study there?
... but you didn't answer my question, and I have yours.
I want to call your post "low-effort propaganda", but it falls short even of that. Like you're running around trying to tell people that clouds are made of marshmallows or something.
a lot of his "burn it down, fuck em" attitude stems from that.
he is, however, completely and utterly narcissistic and greedy, and smart rich people are able to play him like a fiddle.
the KGB picked up on that in, like, the 80's, after Ivana's dad -- an informant for the Czechoslovakian secret police -- passed on details of their relationship to the Czech internal security
Too bad they died off, those were the captains of industry in the 20th century, it was tragic to see them go.
Now it's even more tragic with huddles built from captains of non-industry, "monetizing" rather than producing.
If a consumer recovery is going to happen, what is it being produced that the bottom 10% is going to consume where they can get their money's worth if nothing in the affordable tier qualifies any more? A greater number of people across the board need to get their money's worth on almost everything like never before seen during so many peoples' lifetime at this point.
I liked it better when Trump-like character was almost universally recognized as nothing other than a mindless financial squanderer wrapped around an inherently dishonest core. This was voiced most strongly especially among the rich and powerful.
They used to know better, but that was then and these are not them.
Which has nothing to do with Amazon. Except for the way the political party which leveraged Reaganomics in the 20th century to the effect that average consumers would eventually no longer be able to afford anything other than Chinese-made products, turned out to be the same party which fell behind Trump in the 21st century as if his financial acumen was't completely non-advanced.
And just as consumer prices of the Chinese mass-produced goods continued to fall as desperately needed by consumers due to direct purchase from China, Trump halts that so there's more need to go through Amazon, they can maintain more control that way.
Do the rich hate him because he exploits the working class? Because they think the way he got rich is obscene? No! It's because he won't shut up, and says the quiet parts out loud. They only hate his personality. You would know if they hated his policies, which they don't, because he'd be out on his senile ass before you could say dementia.
good and bad isn't what Meta and AWS is running with. If AWS trains willy nilly on user data like Meta does, AWS is toast. AWS is already taking a huge risk by now using their own take of AWS FDEs competing against their established ecosystem of consulting companies that used to recommend all 583 of AWS products to host a static webpage. Turning of more customers by training on their data is too much of a risk. That's the "good" - a risk analysis.
Meta doesn't care if you stop using Facebook, Whatsapp etc. Frankly, if you are embedded in the Facebook, Whatsapp ecosystem, you don't have a way out as a normie. Even if you are technically capable, good luck publishing your custom Mattermost build for iOS.
Meta has a significant amount of investment in hardware, originally due to Metaverse. The way to monetize that investment is inference and training. That's a do or die decision for Meta. Meta isn't releasing open-weight models because they are "good".
Meta is releasing open-weight models for the same reason Alibaba, NVIDIA or IBM does. Or Google releases Gemma.
There's no "good" or "bad" - the moment Alibaba, NVIDIA, IBM or Google's objective functions change, "Do no evil" will quickly morph into "Do as much evil as necessary without the costs of getting caught exceeding E(outcome)"
Metaverse was 4-5 years ago already wasn't it? That hardware is verging on obsolete.
I don’t condone the use of pirated content for training. But anything that was bought is above the board as far as I’m concerned. Can let the lawyers fight out the minutia and inevitable licensing specifics after the fact (book publishers are hawks).
Fun fact, did you know that every library in the world has an independent contract with every major book publisher?
In cases I am familiar with, they added a condition that a physical book can only be lent X times. After that, the library must purchase a new copy (even if the original copy is in pristine condition). Yeah, hawks.
Public datasets that are liberal with licensed access is one thing, but the best frontier labs clearly went well beyond that.
|----------------|------------------------------:|--------------------:|
| United States | 850,000 | 750,000–950,000 |
| India | 115,000 | 100,000–125,000 |
| United Kingdom | 68,000 | 55,000–75,000 |
| Canada | 40,000 | 30,000–45,000 |
| Germany | 36,000 | 30,000–40,000 |
| Spain | 27,000 | 23,000–30,000 |
| France | 23,000 | 19,000–25,000 |
| Italy | 17,000 | 14,000–19,000 |
| Poland | 17,000 | 14,000–19,000 |
| Japan | 13,000 | 11,000–16,000 |
| Mexico | 13,000 | 10,000–15,000 |
| Australia | 6,000 | 4,500–8,500 |
| Ireland | 6,000 | 5,500–6,500 |
| Singapore | 2,400 | 2,000–2,500 |
That's a lot of "nobody"
Not an employee, just have a "shit ton" of Amazon stock that has been beating inflation very well. The last few quarters of volatility has allowed me to buy and sell, often fully covering my monthly expenses. I'm very greatful for that. Far from a company "nobody" wants to work for. You have a better performing company than Amazon to recommend that I can use instead? I'm all ears and I'll even give you a kickback for your time!
or are they warehouse and package monkeys and logistics?
800k warehouse workers doesn't mean shit w/r/t AI development
They're a good data point on where the robotics labor frontier is.
like pee bottles.
00000010: 626f 74
Amazons interviews were so formulaic and it’s interviewers so robotic it was almost sickening.
Human experience means nothing. Anything that does have meaning gets encoded in the system for the mindless humans to implement.
That’s my take on the view from the top, based on my time working at AWS.
.. and then got let go at 22 months in favour of the AI budget.
Gotta respect the hustle.
HR swore it wasn’t on purpose and I was somehow measuring pay “wrong” by looking at money in my pocket over time; but knowing their HR, I’m sure the valleys in transition from signing bonus to vesting are intentional.
That’s because people are complicated and do things for many reasons, so people quitting when we cut their pay might not tell us they’re impatient.
I suspect the intent is what you say, but I don’t think the outcome is.
I could be wrong because I haven't paid much attention but Amazon has just fast followed everyone else innovation for a awhile so it's in like one ecosystem yeah?
I'm not trying to be snarky, I just haven't thought about aws as having any innovation for a while and I always figured it was because ruthless companies that churn through people as disposable commodities generally don't do tech innovation.
I'm talking on the AWS side, on the logistics side they clearly are doing insane things on supply chain perf optimization.
Generally speaking, on AWS you can buy compute, storage, network and other basic 'goods' and just like a supermarket where you can get either flour and eggs, a cake mix or a ready-made cake, AWS is willing to serve it to you in various levels of ready-ness, and you can start at the level you want, and build your own layers on top if you don't like AWS's solutions.
And AWS's solutions are generally built like that as well, there's no magic to Lambda or Fargate or EKS, its just they run the same EC2s, but with some proprietary magic that lifts some complexity from your shoulders, at the cost of generally some markup, and loss of some otherwise desirable features.
But the core thing is that AWS's offerings are generally solid, so as AI evolves and companies become more willing to take on more complexity, they will move to lower level services, but not necessarily out of AWS itself.
A lot of innovation at AWS is invisible on the outside. They pioneered hyperscale datacenters with publicly accessible capacity, software-defined networks, secure VMs, and so on.
> I'm not trying to be snarky, I just haven't thought about aws as having any innovation for a while and I always figured it was because ruthless companies that churn through people as disposable commodities generally don't do tech innovation.
I worked at AWS before COVID. It was really a varied experience, but it was not _bad_ overall. We prided ourselves on making sure our APIs are solid, stable, and backwards compatible. When we had to make a backwards-incompatible change, we actually contacted every affected customer and ensured that their workflows wouldn't be broken.
Let someone else take on the research and training costs, and charge a massive premium for your hardware.
Data center hardware, by contrast, is a scarce commodity with rapidly growing demand, and therefore likely to capture most of the profits from AI and software development, if current market conditions persist.
Today's IT business trends represent a remarkable change in the industry compared to the situation 50 years ago, when startups like Microsoft and Oracle, whose business models were based exclusively on software development, challenged and overthrew the dominant player, IBM, whose primary source of revenue was from hardware.
the linked article is from July, it is almost three months old.
Now there may be a very relevant recent story; it has the potential to reduce the flow of new open weight models. This could change quite a few expectations, which are based on open weight models:
"China’s STAR 50 Index Falls 40%, Erasing $480 Billion as AI Stocks Crash"
https://www.technology.org/2026/10/27/china-ai-stock-bubble-...
https://www.reuters.com/world/china/china-ai-bubble-has-alre...
Reuters says there is a run, but also deeper reasons:
"There has been no obvious selling trigger in China, although traders said high prices, tax bills and the threat of the US or other nations banning Chinese equipment in data centers all contributed."
So far I don't see any discussions of this story on HN, interestingly.
Sounds like the opposite of what you're suggesting they're doing?
The problem was rather that Microsoft had no real plan what they wanted to get out of AI. If they had a plan, they could have laser-focused their all-in on AI.
Instead, Microsoft just AI-slopped a lot of their bread-and-butter products, making them much worse than before. If AI is as revolutionizing and beloved by users as Microsoft believes, Microsoft could easily have created an offering like "Windows AI Pack" or "Microsoft AI" that should sell like hot cakes.
On the other hand, if you have to force AI functionality onto the users, it should be considered as a strong sign that perhaps it is not something that the users want.
In consideration how much spying functionality was added with Windows 10 (which worsened even more with Windows 11), and that Microsoft made it insanely hard to install Windows 11 without creating some Microsoft account, it is actually very sensible that people don't want to upgrade to Windows 10/11, despite the technological improvements that these versions provide.
The bulk of the hands-on-keyboard time for those disciplines now seems to be using Copilot 365 as "fancy autocomplete" and "make this presentation better", Copilot Notebooks for deeper work, and coming up with all sorts of Copilot integrations and agents for managing email, Teams meetings and chats.
It's almost as if "copilot" has already become the generic term for "professional LLM", and many genuinely can't understand why us techies find it so limiting.
Do they, actually?
3 of the 4 orgs I consult with have actively warned people not to use it, and/or to only use their approved AI tool (OpenAI/Claude/Gemini)
clippy 2.0 is not selling well
Windows (and increasingly Apple) updates are more of a sigh and lip biting experience than an exciting one.
I'm really happy not to have touched Windows recently at all.
At work Microsoft's updates are buggy - Outlook and Teams are pretty awful on corporate Macs. Teams especially, it has a bunch of papercut type annoyances. And updates torture Corp IT.
Notice that you also see Google really struggling (and falling behind) in the AI race. This, despite massive resources and being responsible for key pioneering research. Same reason why IBM has failed in this area. The big established companies tend to shut their brightest people into research departments where they write papers and produce demos while the main cash cow areas remain undisturbed, so they have collections of bright people and they have money, but they rarely let their brightest people make big bets with the money, whereas startups do.
It's also why poor people spend more on lottery tickets than rich people.
Basically you want to run an LLM, Amazon. You want to build an LLM not Amazon
We all feel more productive. But did your rust rewrite or my fancy react dashboard produce more value?
Seriously asking, why would this look like a good business model to be in? If they don't need anyone besides a technical non-developer to create the product, why would this look like a good business? If it's B2B, then most of their customers will have that type of IT person on staff.
Turns out there are ways to monetize commodities, to the tune of double-digit billions in revenue per quarter.
So mighty Google can't manage it. But it's a commodity that's easily copied.
Somehow HN can't reconcile the contradiction even though it's laughably obvious.
Hey anybody could go MVP a ride hailing app over the weekend. It's a commodity app. Should be super easy with LLMs.
Oh shit, wait, you mean I'd have to take a billion users away from GPT and getting those users is now 50x harder than it was four years ago? You mean it's drastically harder to peel users away from existing services than if you get to start fresh? Ohhhh.
They have gigantic moats and so long as they don't do anything too stupid, those moats will just get stronger with time.
No VC is giving you money to try to run down OpenAI and Anthropic. Nobody is writing that $100 billion check. It's already over, HN is just in its standard denial phase.
So you could argue their moat is US government, not their model. I don't disagree with you, just saying what is their real moat.
The Chinese models have no great branding with the masses, they have no service layer that any average user understands. Where does Joe go to use the Chinese model that he doesn't even know the name of? Joe has no clue what the difference between DeepSeek and Kimi is, and Joe will never know that.
They all know ChatGPT, they have all been trained on how to use it and what it is. The inertia is enormous and HNers endlessly undercount how that works because deep tech users all want to think the masses view tools like they do and approach tech like they do (which is entirely false).
OpenAI's future is as an ad platform in terms of its business. That's at least 2x-4x larger as a business opportunity than anything they have going on now. Which is to say $80-$160 billion minimum. That requires they maintain the giant userbase, we'll see how that goes. Losing it requires someone take it from them. Joe isn't just going to wake up tomorrow and go seek out how to use Kimi.
The LLM space is worth a minimum of half a trillion dollars per year in ad revenue. It will subsume all of search. Search becomes merely a service of an LLM, which will perform the function on your behalf. Google is on top of that shift, so OpenAI has its work cut out for it competing with Google there.
Of the US big three, Anthropic is by far the one most threatened by the persistence of the Chinese models. Anthropic will probably never stop being chased by them and will have to endlessly race to try to stay ahead. Anthropic has to figure out how to build out the LLM+ equivalent of AWS before China figures out how to beat them, ie build a very deep service / tool moat around Claude that the userbase gets stuck to. In short dramatically increase the difficulty in peeling away users, it has to become a big chore to switch (which eventually this is exactly what someone will figure out in that area of LLM competition).
For decades, velocity, time to market and features all drove the industry. But AI puts speed in everyone's hands.
But quality is still hard. Maybe harder than ever.
Separate question: "now that AI is here and development isn't a moat any more, how does a business defend itself from competition?"
That second question is a direct result of AI. Five years ago the amount of cost and effort spent on software development was a moat. Now it's not. So that answers the first question completely, I think.
As for the second question - market sector knowledge, customer contacts and industry reputation, all the usual non-technical things that make a moat. Also, we may now be living in a world where software companies cannot expect to be monopolies in their markets any more. This company will have to compete with many similar competitors for market share, like in most other markets. I think this is a good thing. Less enshittification will ensue.
Value was destroyed.
The hope is that all these unemployed engineers can somehow generate value? Even if they do they need to generate enough value that would employ 4 engineers.
If a business can create software cheaper, then it can charge its customers less for that software (or, conversely, if a business can use an AI to create bespoke software for cheaper than it would cost to buy that software) then the customer business saves money. It, in turn, can use that money to improve its offering to its customers. They, in turn, and so on.
Having cheaper things means everyone gets more things. The value created is created further down the supply chain.
[0] https://en.wikipedia.org/wiki/Parable_of_the_broken_window
Where will people work on if both blue and white collar jobs go away? In brothels?
I loved the article that was on HN a while back (original article here [0] be warned it's a quite a long read). It compared the current revolution we're undergoing to the Mercantilist revolution when we stopped being farm workers and started doing other things instead, driven by technological change in agriculture.
The answer they came up with: we don't know yet. Much like the move from subsistence farmers who started working in a factory had to rely on other people to feed them, we have to answer this question ourselves and it'll probably involve doing something very different. And wonderful - subsistence farming is a shit life (despite all the influencers saying how amazing being in touch with the land is) and we would all prefer to be working in an nice clean office than trudging knee-deep in muck in the rain trying to keep our crops alive and pest-free. The same will probably be true once the dust settles on this revolution and we'll all be profoundly glad that AI came along.
The big problem we have to face as a society is that we cannot allow all the value from this technology to be captured and retained by the wealthy few. Our ancestors had to fight the same conflict with factory owners during the Industrial Revolution, and they didn't manage to win that war completely despite all the blood shed and battles won.
[0] https://contraptions.venkateshrao.com/p/cognitive-mercantili...
I can't think of any, I'm sure there are some, but I would really like to know. I just don't think for most business the software is so much of a moat unless it's very complex.
[0] https://www.afr.com/technology/ai-can-make-free-copies-of-ad...
What exactly will cause that? Hardware costs are (generally) going down, and open weight models are getting smarter. The subsidised subscription prices will go away, sure, but I don't see the average token getting more expensive.
Watch what the open weight models can do. Once they can manage your codebase, you’re safe from gouging.
I am often told humans make mistakes, too. But this is an elementary logical fallacy because the type, frequency and severity of the problems is never the same.
Also, humans are more likely to do as told, whereas AI often takes liberties (i compiled a book from a github repo and it did things on its own in the book though i had explicitly forbidden that).
And it's measurable. We don't just feel more productive; we get grants, we build systems, and we make plans we simply couldn't have before.
Maybe you could use an AI assistant to help you write better HN comments.
i add more value to my company because it lets me ship much faster, debug issues faster, onboard new people faster
total enterprise ai spend is climbing https://ramp.com/data/ai-index#adoption#overall
i know multiple stealth startups with 9 figure revenues and profitable and growing
lab arr is climbing
given the recent crackdown on token spend and from speaking to execs they are very focused on the exact ROI they will get from AI
Note that business revenues in competitive markets may actually not change that much, because if everyone is using the same models there's no competitive advantage from using AI. but if one of them didn't use AI they would fall hopelessly behind
But make no mistake. Being more productive does not equal higher profits because your customers think you've added value. Both value and productivity are highly subjective terms which is why it is measured differently depending on where you work.
Most businesses adapted to AI by shipping a lot more features in smaller increments of time. What they should have done is become much more ambitious and build moon shots in two quarters instead. You need to go big or go home if you're going all in on AI.
ship fast go-to market features in the blink of an eye when it becomes necessary
Necessary go-to market features are useful
It's 1987: The PC is underdelivering.
It's 2000: the Internet is underdelivering.
It's 2010: smartphones are underdelivering.
Even quantized local models like Qwen3.8 are good enough to handle a lot of drudgery. For example, I hooked my wife's computer up to a local model, and now she finally has a clean desktop. She just drops spreadsheets or powerpoints or whatever on her desktop and automation takes care of everything.
The caveat is that I developed software for 30 years, so I understand most of what it produces and can guide it to make fewer mistakes and fix what it can't do. The biggest challenge has been keeping my requirements focused - it's very easy to gold plate everything and prematurely optimize.
The biggest question I ask myself is why am I even still doing this. The smart people are probably already trying to figure out how to pipe raw matter and energy to their homes to build replicators.
The future is already here, it's just not evenly distributed.
You don't want to be in the CD selling or radio business with the Internet & digital & streaming & smartphones & iPods bearing down on you. Go for concerts. Notice that humans are spending more than ever on live concerts. Disney World and Disneyland still command extraordinary pricing power, despite everybody hiding inside on their phones.
LLM clones of eg Photoshop will end up being worthless, like ideas generally. The software will be worthless for the same reason ideas are: they're too easy. Users are hard, software will be (relatively) easy.
Is the value in McDonalds in the food? Everybody knows it's not. McD captured consumers, their heft is extraordinarily difficult to displace. Ie habits, brand awareness, location, etc.
I doubt that. If a clone out there offers all the Photoshop's features for free or much lower cost, without additional learning curve, then many professionals will switch to save money.
McDonalds would be gone too if somebody opened a restaurant chain serving similar food for free or 80% lower price. But that's impossible to do profitably, while freeware software has existed for a long, long time.
The many professionals are the ones that could most easily understand how to switch to GIMP, which is free to use. They could save enormous amounts of money. They prefer to spend money and stick with Photoshop instead. That has persisted for a very long time now.
What the hell does this even mean? Literally half of the code in operation I the world is still COBOL. The fuck is this nonsense wiped out
Unless you're making the claim that software engineers aren't involved or that only writing code is software development.
Also remember, the majority of all human software development has been wiped out 18 months ago but the dwindling will keep going by the quarter until the majority of all human software development has been wiped out.
One million software developers will lose their jobs in the US over the next ten years.
The big firing waves will begin when valuations decline as the initial AI boom slows. That has started already, although it's early. Companies like CoreWeave will basically collapse. Investors will demand & reward the layoffs, efficiency push, the end of rah rah times will prompt the inflection on employment in tech. And it'll never be like it was ever again.
For 1 million software engineers in the US to lose their job you'd have to shrink the industry by 62.5% and nobody is doing that. We employ more software engineers this year than last year, which was higher than the year before which was higher than a year before. Get off the social media and look at real statistics and news.
This gleeful nihilism isn't health friend
HN is in denial. Just watch what happens over the next ~36 months.
In all likelihood, massive worldwide recession due to combination of oil issues and debt-ai bubble popping. That will be combined with instability due to democracies cracking up, wars popping up faster and faster all the while the richest and most powerful people are increasingly desperate to destroy democracies before it is too late for them.
software is easier to build -> the world builds more software
But yes, the act of translating spec into code as a 'programmer' is gone for good.
There’s no way this is even remotely true. JavaScript alone is probably 90+%, considering it’s on almost every website.
Remember that all of the code in operation includes code like the microcontroller in your mouse, the hundreds of chips in your car doing things like keeping the engine running or "window goes up when I push the button," the DVD menu (which is actually Java code), not to mention all of the staid systems responsible for things like writing your paycheck every other week.
The bottom 2/3 of the human market is entirely dead. Their jobs are gone (AI-shored) they just haven't all gotten the forced retirement or pink slips yet, while they click a few more prompts for a few more pay checks. The frontier pushing forward will knock out most of the remaining top 1/3 without breaking a sweat over the next few years. The really good developers will graduate into babysitting AI as managers.
It is legitimately sad, it's millions of high paying jobs just in the US alone. It's going to be brutal on the housing market. Automation came for another field.
If you have a big mortgage and no savings, it’s going to be tough. Stop taking the Waymo now.
My suggestion to anybody staring this down is: do something for yourself while you can, while we're in the early going. Most people still have no understanding of AI, most tech employees are doing absolutely nothing interesting with it other than prompt prompt prompt prompt while at work and effectively doodling at home.
Go high value and go for the users. Do things that don't scale, that actually rings more true now than ever. Focus on areas of tech where it's really hard to get the users, where you have to practically hand acquire them to get started. Users are the moat.
It's not build the software and they will come. That's what nearly everybody has wrong about what is going on right now. The value is not in the software. Users, users, users, users - they should be your primary focus.
Also consider looking at software more as a layer related to something else you do, not as The Thing you do. Related to another service you offer, improving that service.
Let's say you started consolidating in the physical self-storage space ten years ago. You read that it was a very wide open field, where the market was not yet highly concentrated, 3/4 of the market was mom & pop operators. Over the next ten years you built up a lot of storage locations. LLMs can amplify your business, they can help you automate, layer better software on top of your physical business. You know what LLMs can't do? They can't build storage units, and humans love their storage space. That's just one example out of a zillion out there.
And in before ‘JumpCrissCross chimes in about inference being profitable while ignoring all the investment costs needed to get inference to where it is.
I mean, if anything summarizes where we have gotten to as an industry this quote is sure it. From Uber using VC funds to subsidize growth to where the AI players are at...
but profitability is a very narrow view on the labs, and investors are pricing in future free cash flow
you need to take into account the derivative, what is their growth? their opex and capex? their revenue growth? their moat? New product lines and applications?
Not saying that it's a homerun, but there is a case to be made for the labs being wildly profitable in 4-10 years
Before someone else jumps in talking about "unit economics": unit economics only matters if you can sell the units long enough to make the initial (capitalized) investment (i.e., R&D) back. Models aren't commercially viable long enough for the unit economics to work.
They have been chasing extreme growth with cheap money. That enabled them to hyper scale at a pace the world has rarely seen before. They have made the correct choice so far.
The absolute dumbest thing they could have done up to this point is slow down to try to turn a profit. They have not needed to. Market capture is the sole thing they should be focused on. Here's a hint: how many users does GPT have now versus three years ago?
There appears to be very widespread confusion about what purpose turning a profit serves and how business operations actually work.
Go back in time and tell Uber in ~2013-2015 that it needs to immediately become profitable and not focus on capturing the market.
Uber has to pay back all the billions of dollars it wasted trying to get people to use their platform, and now I can't afford most Uber or Lyft rides, it's way too expensive. So did market capture really work if they had to raise their prices so much as to make the service unworkable for most people? The only time I take uber is if my company pays for it, and that is getting less often.
AI is also going to have to become way too expensive so they can pay back the billions (trillions?) it is borrowing. And, when everyone is laid-off or fired because AI took their job, then who's going to be paying for AI? I guess the companies that lay off all the humans will, but when the wealthy business owners realize they've replaced everything with AI and nobody is buying what they are selling because there's no more jobs, then what?
Sounds like they should worry about profitability, and a lot of other things too.
Hate to say it, but in this case you’re not the market so you are irrelevant.
Here’s data from Chicago, USA, showing Uber and Lyft dominate the market now, so yes, it worked.
https://toddwschneider.com/dashboards/chicago-taxi-ridehaili...
$32b -> $37b -> $44b -> $52b -> $57b
That's Uber's sales growth year by year. And now they're solidly profitable.
The sole threat to Uber is the Waymo / Tesla automation angle. A merger is most likely there.
It feels like a related statement to someone willing to burn their own country to rule over the ashes.
The price of a given point on the intelligence frontier is only going to get cheaper, likely 10-20x over the next few years.
This isn’t the “hallucinate everything” style of model like the earlier GPT models…
Downside is that OpenAI is already having ads on their chat interface, so I guess it will be crap any time soon.
My mother-in-law has a complicated set of medical conditions and sees half a dozen different doctors. Fable with max thinking came up with a bunch of interactions between drugs and conditions that her doctors hadn't considered. We've run its recommendations by one doctor so far and he agreed with everything.
My brother has one friend who used a frontier AI to diagnose his wife's uncommon cancer, and insisted on a test for that. It was correct and she's fine now since they caught it early. Brother has another friend whose wife had back pain after surgery; medical team didn't find what was wrong, he uploaded the scan, and the AI said "oh they left in a staple," which turned out to be correct.
I've started using it to work on a cookbook for the medically prescribed diet I'm on.
Sure it has replaced Google for a lot of people. But was that driven more by Google breaking search and actively ranking down good results and putting blocks in the way of seeing even the bad results?
Google had (have?) people using their service out of habit but built a hostile enough environment. We have this "contradiction" where non-tech folks are supposed to be difficult to move off "what they know" yet here we are where they are using ChatGPT to search and be a doctor.
- File taxes and find issues that "accredited", "certified" entities missed. In one instance, my federal refund increased by nearly $5k. At $20/mo, just the refund is two decades worth of value.
- Appeal health, home and car insurance claims successfully
- Find out details of interesting plants without forking out tens of dollars for an app that claims to do the same but fails to do so most of the time
- Summarize never ending email threads with 20 disinterested parties who would rather not be on the email but have to respond to show they are "engaged"
- Write a convincing email to my mother in law why I cannot meet her this weekend and hold her hand through the tough times she's having with her ethnic neighbor who annoys her with their "ethnic" music and "smell of fish" BBQ
- Locate the user manual for my microwave and what that rapid 3 beeps every 1s for 2 minutes that starts at 10AM since the last 2 days mean
Unless you're specifically meaning microwave ovens, which are of "limited utility" in a full-service kitchen. I'd have to agree there.
My previous microwave oven was so intensely powerful, that I needed to dial down the power to "7/10" so that it would not destroy my frozen meals. I could never run it at 10/10. My current appliance is very nice, and has an automatic button to run until doneness. Works every time now.
2. greatest hacking tool in history
3. greatest corporate espionage tool in history (stealing of Navier-Stokes solution)
4. greatest accountability transfer tool in history (whoops it was the AI that shot those missiles)
5. greatest capital raising tool in history (we use AI in our startup, and you should give us money)
6. greatest scamming tool in history (cloning people voices for scam calls, etc..)
Edit:Whoops, I didn't read "positively"
It's the same with AI. The technology is here now. My friends in financial advice and market research have not lost their jobs immediately, because of momentum.
"democratization of information", maybe if you were too lazy to click on a google link
Like, a normal person is not going to open up an arXiv preprint from Google search because for approximately the entire population, it's going to be way beyond the limits of their comprehension. They won't even glance at the link, if Google even serves it up as a result.
But, with a chatbot, they could actually ask questions and get answers in a format they can understand. They may get a very rough understanding of the topic, but I would bet it's 100 times better than the same reference explained by modern "science journalism," which insists on presenting the most clickbait-y interpretation they can find.
And this is not theoretical, I've seen my high-school level kids figure out undergrad (or higher) level concepts like locality sensitive hashing with the help of ChatGPT, and I know they understood it because they could sketch out working solutions using those concepts. They didn't get the Mathematical underpinnings, but they understood the high-level behavior and that was enough.
Everybody can now understand anything if they really want to. Nobody is doing it today because they don't know it's possible. But they can, and we should encourage them to do so.
It is tools which are the commons (such as FOSS) and tools such as Wikipedia. But also open standards. LLMs cannibalize on these tools.
Furthermore, no customer wants an AI as customer bot. Especially elder people who are a bit computer illiterate can't find the real support (dark patterns). Can AI help a customer support representative? Oh, absolutely. AI can help a skilled person become better/quicker at certain tasks. But it cannot replace them, and it shouldn't either. If we do, we get all these 'fantastic' tools like *claw.
Advances are still coming rapidly -- we have quite a bit of time before this settles.
> Amazon Is Gutting Its AI Division [July]
https://news.ycombinator.com/item?id=49082344
This isn't a blip on anyone's radar.
- Single point of billing and vendor is still king in enterprise - look at how much of Ant's revenue flows through AWS Marketplace
- With OSS, models may move to a more of an AWS-shaped problem: host with 5 9s, provide the infra and shovels, do it as cheap as possible, fit them neatly into rest of AWS stack. I think this is actually a huge risk to OpenAI and Anthropic, as more commodization gives hyperscalers more advantage
- Still owns a huge chunk of Anthropic and leverage over compute
With this in mind, I'm not sure why they need a frontier model vs. the playbook they did with Linux (i.e. Kimi-K3-AWS-Enterprise-Edition). They are kind of playing both sides right now I guess, but could really stir things up if they wanted to.
Strange headline given that it seems they are actually just refocusing resources to train a frontier model?
This outcome for their years of squandered effort honestly doesn't surprise me as, other than the Graviton/Trainium work, the Amazon efforts were scattershot, half-baked and underinvested. Only the Siri team is more pathetic than the Alexa team. Well maybe Co-pilot stands between those two somewhere.
[0] https://nova.amazon.com/act
Amazon makes 60B+ in profit per year, they don't need to tank that by losing it on training AI / selling tokens.
There are plenty of open weight models out there for it to use with most of the benefit and very little of the downside.
I have tried asking people if they've ever used an AI product from certain companies, and you only seem to get "no", or surprise that they have any AI products, for some.
They are just changing their strategy to invest in a single model? It's not like their AI department is in shambles...
as an amazon shareholder, I applaud this.
Source: https://www.businessinsider.com/amazon-overhauls-ai-strategy...
Couple comments then:
https://news.ycombinator.com/item?id=49082344
“So… what happened to the previous team?”
I might try Google's home option again but I'm not a fan of giving Google second chances since they seem to care even less about committing to products.
Amazon is not going to pay for AI internally, and sooner or later they will have to add something internal.
It will always remain weird to me that a publication called Futurism is so outright negative towards our newest technologies.
Values drive behavior. Who owns / runs futurism and what are their values?
Journalism / media as activity needs to be sustainable and unfortunately in our current world it's seems very difficult to sustainably remain high integrity. Sometimes it may be greed, but most of the time I think it's survival.
In this case AI is part of what is threatening journalist survivability, so it kind of makes sense.
If true, Amazon is not really gutting its AI division.
funny story actually. I was recruiting an engineer from another stressful team. I pitched to him that we have about 4-6 sevs a quarter, and I was worried he would balk.
He said "that's nothing, when I was at Amazon we had 30 Sevs a week" .
I didn't feel so bad .
This is something that requires an ethos completely different from how amazon works and the way jeff bezos sees the world.
They already sell a lot of the shovels for training and inference and make ample money from that, and there are enough open weight models on the market to make it infeasible to build a large enough moat.
> Instead, labor and computing power are being diverted toward a singular “frontier-model effort”
It’s a consolidation, not a withdrawal.
> Instead, labor and computing power are being diverted toward a singular “frontier-model effort” led by Pieter Abbeel, director of the Berkeley Robot Learning Lab whose robotics company Covariant was hoovered up by Amazon in 2024. At the time Abbeel joined the tech giant, Covariant’s founders were working on AI models for robots, which lines up pretty well with Amazon’s push to automate everything it possibly can.
Amazon can't hope to compete in the LLM genAI (what is colloquially known as AI anyway) space. That is toast, the race is over for them, too much competition, no way to make money.
But industrial automation? Globally, Amazon employs about 1M people in fulfillment/operations roles, the utter majority of their headcount. If there is one company on this world that has the money and compute capacity to spare for a high performance, real world application of fusion between modern machine learning and robotics, it's Amazon - and to make life easier for Amazon, unlike every other competitor they don't have to pay or partner with another company for real world testing. They can just go and deploy a robot in their warehouse if they want to.
Achieve a breakthrough in a fully automated picker robot, and they can save half a million people on their payroll, if not more, they'd save shrink loss and handling mishaps as well. That is one hell of an opportunity to save money by R&D. And if they patent everything they do, they can lock their competitors out of that technology for a loooong time.
you want edge inference, from decision to small models.
That's it. It will be 2c, please.
You'd think with all the data collected by these tech companies, they'd understand the common usage scenarios of their products. But instead we have AI Alexa that can't egg timer, or Gemini Assistant that can't navigate or make phone calls.
Latency is the #1 UX killer on voice apps, and the latency was terrible compared to Alexa. I used it for like 15 minutes and reverted.
All anyone was talking about was how much worse it was for common tasks.
The title is clickbait - Amazon is redirecting resources toward another major AI project and that does not, by itself, establish an overall "retreat from AI". The last 5 years have been a time of flux and now companies are starting to focus on niches and gaps in the market.
AI takes substantial focus and investment and we are compute constrained - AWS is correctly consolidating and refocusing by scaling back several of its existing Nova projects (Canvas, Reel) and concentrating resources on a new flagship model effort led by Pieter Abbeel.
If the claim is "sustained failure", what failed against which objectives? Technical capabilities? customer adoption? economics? delivery? The article does little to establish any. Consumer name recognition is a weak substitute for evidence about an enterprise model business!
A good report would have included benchmarks, revenue, or profitability evidence supporting the article's broad failure claim - this lazy "vibecoded" report doesn't even come near.
It credits Bloomberg while linking to Business Insider. That is a straightforward sourcing error - a journalism 101 no no.
If this is the new generation of shoddy, sensationalist "reporting", good riddance!
Please don't accuse people of writing AI-generated comments. If you suspect it, please email us: hn@ycombinator.com.
it doesn't concern you.
If someone started from zero, how long do we think it would take for them to reach the cost/capability front currently provided by gpt-6.1-terra? How long until OAI figures out a way to distill that down again for pennies on the dollar?
Apple seems to have this sorted. You won't find them chasing OAI, Google and Anthropic around the playground.
https://blog.google/company-news/inside-google/company-annou...