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

The Generative AI Con

wheresyoured.at

Post reply on HN
Print view View original

171–180 of 503 posts

Re: The Generative AI Con

#171

Post by stpedgwdgfhgdd »

Software development to me has always been about the 80-20 rule. You build 80% of the functionality in 20% of the time. Next you spent 80% of your time to build the remaining 20%.

With LLMs it feels we are getting near to 90-10. Finding the bug in those good-looking pieces of generated code is pretty hard. (After all, you did not pay a lot of attention to the generated code, it looked pretty solid) Some will argue that the LLM should spot the bug, Indeed, it should ask clarifications about the requirements. One day… but you need an expert to understand and answer the questions for that last 10%.

Re: The Generative AI Con

#172

Post by ben_w »

trhway wrote: post #124
automatic6131 wrote: post #106

Earlier quoted context omitted.

>There is not much need for those trillions in return Whilst you are correct that big tech cos do not need the return to survive, that's not how public markets work at all, and thus not how the incentives for those in charge of the companies work, and so making you actually wrong.

If i were wrong, those companies would be distributing that cash to shareholders instead of chasing any promise of any big chance. If investment in AI don't pan out (i do think that it will pan out, and those trillions will come) then those companies would just pour even more billions into whatever big thing/promise would come next. Rinse and repeat. Because some of those things do generate tremendous returns, and th…

Markets are funny things.

US right now is run by someone whose explicit promises, if actually implemented, have an obvious immedidiate 13-14% reduction in GDP — literally, never mind side effects, I'm not counting any businesses losing confidence in the idea that America is a place to invest, this is just direct impact.

DOGE + deportation by themselves do most of that percentage. The tariffs are a rounding error in comparison, but still bad on the kind of scale that gets normal politicians kicked out.

And yet, the markets are up.

Re: The Generative AI Con

#173

Post by simne »

I wonder, when people will begin read books (or at least learn documentation).

> OpenAI burned more than $5 billion last year.

Well, this is semi-true. When speaking about LLM technology, must be honest, and make difference of base (or foundation) model training, vs fine-tune it for purpose.

Sure, if you just use base model, you also could gain some profit, but real value of LLM achievable if you got already done base model and fine-tune it on your target task.

What this mean - base LLM are just learn language structure from really huge dataset (for example, entire Wikipedia), and this is really expensive, but when you fine-tune LLM from for example, your corporation product documentation, it will become AI-consultant about your corporation. Or you could fine-tune LLM from children story book, and it could indefinitely generate texts similar to that story. BTW, rumors said, some orgs fine-tuned GPT-3 on their company codebase and have very interesting results on code generation (much better than with base model).

Fact, base model training really cost millions (Llama-2 official cost $5 millions, and I believe it much more than claims of Chinese about deepseek R1 cost also $5 millions).

But fine-tune GPT-4o now cost about 20 bucks for 1 million tokens, and inference is $3.75 per million input tokens and $15 per million output tokens. For GPT-4o mini, training cost is $3 per million tokens, and inference is $0.30 per million input tokens and $1.20 per million output tokens (from official announce on OpenAI developer community).

If you consider fine-tuning of GPT-3 class model (or for example, similar open source model), official prices are just few bucks for million tokens (run it on your own infrastructure will be slightly more expensive), which I think very tolerable and already affordable for small companies.

And I admit, just few Billions of market is not scale of big thing, but I think, it is just because conservative corporate tops, and because security problems of current implementations, and will change nearest years.

Re: The Generative AI Con

#174

Post by andsoitis »

suraci wrote: post #78
adpirz wrote: post #9

Earlier quoted context omitted.

The "iPhone moment" gets used a lot, but maybe it's more analogous to the early internet: we have the basics, but we're still learning what we can do with this new protocol and building the infrastructure around it to be truly useful. And as you've pointed out, our "bandwidth" is increasing exponentially at the same time. If nothing else, my workflows as a software developer have changed significantly in these past t…

> If nothing else, my workflows as a software developer have changed significantly in these past two years with just what's available today, and there is so much work going into making that workflow far more productive. this is exactly the problem The more productivity AI brings to workers, the fewer employees employers need to hire, the less salary employers need to pay, and the less money workers have for consumpti…

Less productivity seems like a worse path.

Re: The Generative AI Con

#175

Post by porridgeraisin »

sambapa wrote: post #67

I always wonder how LLMs will achieve superintelligence when they are, by definition, average.

To this pedantic point, If the average written intelligence of all humans alive and dead is > the max intelligence of all live humans who are also willing/positioned to do the same task at the same time and at the same place.

But yeah, I don't think LLMs (the current core architecture) can provide super intelligence. I think it needs a bit more than next token prediction architecturally speaking.

Re: The Generative AI Con

#176

Post by jcgrillo »

jsnell wrote: post #159
jcgrillo wrote: post #140

Earlier quoted context omitted.

> Cost as in, cost to you? Or cost to serve? This. IIUC to serve an LLM is to perform an O(n^2) computation on the model weights for every single character of user input. These models are 40+GB so that means I need to provision about 40GB RAM per concurrent user and perform hundreds of TB worth of computations per query. How much would I have to charge for this? Are there any products where the users would actually g…

> IIUC to serve an LLM is to perform an O(n^2) computation on the model weights for every single character of user input. The computations are not O(n^2) in terms of model weights (parameters), but linear. If it were quadratic, the number would be ludicrously large. Like, "it'll take thousands of years to process a single token" large. (The classic transformers are quadratic on the context length, but that's a much s…

Ah got it, that's more sensible. So is anyone making money with these things yet?

Re: The Generative AI Con

#177

Post by simonw »

suraci wrote: post #78
adpirz wrote: post #9

Earlier quoted context omitted.

The "iPhone moment" gets used a lot, but maybe it's more analogous to the early internet: we have the basics, but we're still learning what we can do with this new protocol and building the infrastructure around it to be truly useful. And as you've pointed out, our "bandwidth" is increasing exponentially at the same time. If nothing else, my workflows as a software developer have changed significantly in these past t…

> If nothing else, my workflows as a software developer have changed significantly in these past two years with just what's available today, and there is so much work going into making that workflow far more productive. this is exactly the problem The more productivity AI brings to workers, the fewer employees employers need to hire, the less salary employers need to pay, and the less money workers have for consumpti…

What's your opinion on the productivity boost open source libraries have brought to developers?

Did all of that free code reduce demand for developers? If not, why not?

Re: The Generative AI Con

#178

Post by nbuujocjut »

GitHub Copilot and similar tools make good developers more productive. This alone is a genuine use case with some associated value. Is it enough to justify the valuations of OpenAI etc? Probably not by itself. But I expect other industries have similar productivity boosts where people learn to use the tools appropriately.

What’s the total business opportunity of making all knowledge workers 10% more productive (to pick a more modest goal than outright replacement)?

Re: The Generative AI Con

#179

Post by baggachipz »

I had to write a similar post[1] a few months ago because I was so tired of everybody I know telling me we're all gonna be without jobs and we're about to enter a new epoch. I'm so sick of the hype and that will be the real thing that dooms us all.

[1] https://blog.curtii.com/blog/posts/the-laypersons-guide-to-a...

Re: The Generative AI Con

#180

Post by Red_Comet_88 »

"We are in the midst of a group delusion — a consequence of an economy ruled by people that do not participate in labor of any kind outside of sending and receiving emails and going to lunches that last several hours — where the people with the money do not understand or care about human beings."

Regardless of the hostile tone of the article, this stuck out to me as an incredibly poignant description of the current tech/finance elites' mindset.

As most of us who have tried LLMs can attest, they are indeed stochastic parrots with no capacity for knowledge or understanding. This is best exemplified by their non-deterministic outputs, wherein they give different answers to the same question if asked enough times. This is not how a human brain works. Perhaps it is a small building block, but the systemic architecture required to reach brain level is currently not in sight based on what I'm seeing.

Post reply on HN