Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Hater's Guide to the AI Bubble

wheresyoured.at

Post reply on HN
Print view View original

141–150 of 168 posts

Re: The Hater's Guide to the AI Bubble

#141

Post by 0x000xca0xfe »

frithsun wrote: post #31

I believe this is a "good" bubble in the sense that the 19th century railroad bubble and original dot com bubble both ended up invested in infrastructure that created immense value. That said, all of these LLMs are interchangeable, there are no moats, and the profit will almost entirely be in the "last mile," in local subject matter experts applying this technology to their bespoke business processes.

I hope it will bring us reliably good memory bandwidth in consumer devices, an area where many hardware vendors are skimping.

Re: The Hater's Guide to the AI Bubble

#142

Post by snapcaster »

urbandw311er wrote: post #117
snapcaster wrote: post #99

Earlier quoted context omitted.

who you going to believe? random studies or your own lying eyes?

You say “random studies” like you’re trying to discredit them in some way. But they’re not “random”, they’re controlled, documented, objective and published for you to go and read the results.

personal experience > some p-hacking academic

Re: The Hater's Guide to the AI Bubble

#143

Post by K0balt »

tim333 wrote: post #130
K0balt wrote: post #86

I too am deeply skeptical of the current economic allocation, but it’s typical of frontier expansions in general. Somehow, in AI, people lost sight of the fact that transformer architecture AI is a fundamentally extractive process for identifying and mining the semantic relationships in large data sets. Because human cultural data contains a huge amount of inferred information not overtly apparent in the data set, ma…

Different algorithms do different things but “generative” AI can certainly come up with new stories and images and with different algorithms AI can work with not fully solved problems like protein folding.

Coming up with new arrangements of bits is not a particularly hard problem on its own, but the current crop of ai is certainly able to do that in the extractive process, in fact randomness is a key part of training and inference. But making new things from old parts does not constitute innovation, insofar as the arrangements follow known paths.

That doesn’t make it non useful. It just makes it non innovative.

Trial and error within a defined problem space is an area where automation can definitely be useful. Once again though, the result is not innovation but rather automation of labor.

There is a -lot- of labor requiring mind numbing repetition or iteration. The vast majority of labor falls into this category, and exists in fundamentally solved problem spaces, but still is complex enough that the algorithms involved are opaque. This is where the current type of AI can work miracles when trained with enough oblique data.

Re: The Hater's Guide to the AI Bubble

#144

Post by troupo »

tim333 wrote: post #137
troupo wrote: post #135

Earlier quoted context omitted.

That's a huge if that you can only accept if you buy into "AGI has been achieved internally" marketing.

I'm thinking more that AGI will happen within a few years. It's one of the reasons for the present financial weirdness. Throwing so much money into current AI would make no sense if it stays as it is. It only makes sense if you think the tech will improve.

You say that arguments for AI and crypto are somewhat distinguishable while presenting arguments indistinguishable from crypto.

Billions of dollars were spent on crypto as well.

Re: The Hater's Guide to the AI Bubble

#145

Post by appreciatorBus »

bandrami wrote: post #133
appreciatorBus wrote: post #108

Earlier quoted context omitted.

It's also a common misconception in housing debates! "B-but the developer always has to make the money back so rents & prices can never go down!" "That's the fun part, they don't!" The builder/buyer/lender/landlord/etc can go bankrupt but as long as the building actually got built, it will carry on and benefit the rest of us, regardless of what happened to the ppl who paid for it to be built. Also fun when landlords…

IDK, there are whole developments in FL that are just going to be bulldozed. Same thing happened in TX after the S&L crash 35 years ago.

I’m not familiar with either, but feel free to share a link.

If the buildings, I assume single family houses, were built in places that few people want to live, or places people are simply forced to live because they can’t live where they want to live, then it wouldn’t shock me if someone chose the bulldozer as the best option. I’d argue that the root cause is bad land use law that doesn’t allow construction in the places people really want to live, but that’s a whole other topic of course :)

My argument, both for AI, railroads, telecom and construction, is that if you built something tangible and useful, it can outlive your financial arrangements, and go on to serve humanity, even if you go belly up.

But for residential construction, location is everything, you could build a theoretically useful building, but in a location that few people want to voluntarily live in, far from jobs services, shops, friends, and family, etc. Ofc, you could build a railroad between two unpopular destinations, or run a fibre optic line between two places with little demand, and those would probably not outlive your finances and might be torn up or abandoned too.

Re: The Hater's Guide to the AI Bubble

#146

Post by UltraLutra »

hotpotat wrote: post #69

Lots of in-depth analysis, but I think the author is very clearly emotionally invested to the point that they are only drawing conclusions that justify and support their emotions. I agree that we’re in a bubble in the sense that a lot of these companies will go bankrupt, but it won’t be Google or Anthropic (unless Google makes a model that’s an order of magnitude better or order of magnitude cheaper with capability p…

> Claude is simply too good at coding in well-represented languages like Python and Typescript to not pay hundreds of dollars a month for (if not thousands, subsidized by employers).

I think the cost is more in thousands to cover inference. And, no, I don’t think it’s been proven out that an engineer is so much more productive to justify thousands of dollars a month cost. The models are great for greenfield projects. But a lot of engineering is iterating and maintaining an existing code base——a code base that the engineer is fluent in. So the time savings is writing code specific enough to implement a new feature vs writing a prompt specific enough that the AI can write code specific enough to implement a new feature. The difference between those two tasks is the time savings.

Say that difference is like 10%. You save 10% of your time by using AI, meaning you have 4 more hours a week than you did before. Are you going to spend 4 more hours writing code? No. Some will be spent in meetings. Some will be spent reading Hacker News. Maybe you’ll get two hours a week of additional coding time. So you’re really only increasing your output by 5%.

The so the employer gets 5% more from you if you have AI. If your salary is 10k per month, they wouldn’t pay more than $500. Per month. And you’re probably costing Anthropic >$10k in inference costs per _week_. The economics just don’t make sense.

You can sub out the numbers here and play around with the scenario. I think the cost of inference needs to drastically fall. And I don’t think that happens soon. What might happen 10 years from now is developers are given a laptop with a built-in GPU for AI inference that does much better code auto-complete using AI. That’s something an employer can pay 3k-5k for _once_ as a hardware investment. But the future of AI coding won’t be agents. It won’t be prompt-engineering. The models aren’t going to get much better. It will be simple and standard and useful but unimpressive. It’s going to feel boring. It’s going to feel boring. When it’s working, when it’s mature, when it becomes economical, it always feels boring. And that’s a good thing.

Re: The Hater's Guide to the AI Bubble

#147

Post by xela79 »

zild3d wrote: post #94
xela79 wrote: post #41

make a technology very affordable, get people hooked. Then when LLM have basically destroyed the open web, charge more for accessing and searching that wealth of human created knowledge. Profit $$$ Ethical approach? hell no. What do you expect from an unregulated capitalistic system.

> What do you expect from an unregulated capitalistic system. Competition, fortunately

> Competition, fortunately

so there's no competition when there are no rules and regulations... ? interesting.

all those sports without rules or regulations, like american football where anything goes.

Re: The Hater's Guide to the AI Bubble

#148

Post by UltraLutra »

UltraLutra wrote: post #146
hotpotat wrote: post #69

Lots of in-depth analysis, but I think the author is very clearly emotionally invested to the point that they are only drawing conclusions that justify and support their emotions. I agree that we’re in a bubble in the sense that a lot of these companies will go bankrupt, but it won’t be Google or Anthropic (unless Google makes a model that’s an order of magnitude better or order of magnitude cheaper with capability p…

> Claude is simply too good at coding in well-represented languages like Python and Typescript to not pay hundreds of dollars a month for (if not thousands, subsidized by employers). I think the cost is more in thousands to cover inference. And, no, I don’t think it’s been proven out that an engineer is so much more productive to justify thousands of dollars a month cost. The models are great for greenfield projects.…

* And you’re probably costing Anthropic >$1k in inference costs per _week_. Not >$10k. Typo.

Re: The Hater's Guide to the AI Bubble

#149

Post by 827a »

tim333 wrote: post #134
827a wrote: post #120

Earlier quoted context omitted.

The nasdaq composite p/e at its peak during the dotcom bubble breached 200. Today we're at 40, and Nvidia alone is at 49. As much as everyone wants this to be a bubble: it isn't. ChatGPT was the fastest "thing" in history to reach 100M MAUs, and is believed to be a top 5 most visited website today, across the entire internet. Cursor was the fastest company in human history to reach $500M in revenue. Midjourney, the c…

>we physically cannot buy enough GPUs to satisfy demand is caused by mispricing - VC money is used to pay for the GPUs but the product is mostly given away for free. If companies just charged the public what the service cost to provide usage would go down dramatically.

If they had to actually fund R&D and footprint expansion with customer money: Absolutely yes. But, that's exactly what VC money is for. If VC money can cover expansion while customer money can cover per-token incremental costs, eventually they won't need to expand quite as much, and then profitability should catch up.

In other words: How literally every tech business that has ever worked has worked. This isn't news. This isn't novel. This is just how it works.

If you want me to feel fright that the world is coming to an end, wake me up when Google (one of the world's largest frontier AI labs by any measure) isn't posting $35B in profit on a 39% gross margin every quarter, or when Meta (who is reported to be paying AI researchers nine figure comp packages) isn't making $16B on 40%. The amount of money these companies make is disgusting, its so disgusting that they can blow a hundred billion on GPUs and key people, they could write it all to zero two years later, and its all a teeny tiny little blip on their graphs, it becomes the third line-item in their quarterly board meetings, behind far more important stuff. The reason why some of you get so freaked out is because you literally cannot comprehend the scale these companies operate at, and how financialized their operations are.

Re: The Hater's Guide to the AI Bubble

#150

Post by falcor84 »

queenkjuul wrote: post #127
tim333 wrote: post #106

Earlier quoted context omitted.

Although if there is some small possibility that X will be replaced in a couple of years shouldn't people be able to consider it?

Small possibility some natural disaster ends all life on earth in two years. Large possibility it will within decades But hey more fun to pretend the chatbot will turn into Terminator

Well... whether the machines take over or just give us a common enemy, at this stage, the rise of Skynet seems to me like the highest likelihood path for global coordination.
Post reply on HN