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

AI is slowing down

wheresyoured.at

Post reply on HN
Print view View original

661–670 of 820 posts

Re: AI is slowing down

#661

Post by oblio »

simianwords wrote: post #654
emil-lp wrote: post #643

Earlier quoted context omitted.

I didn't find the answer there, that's why I asked. What hardware is needed, how much of it, cooling, and what does it all cost you? Or are you saying I can take my old desktop and serve Deepseek v3.2 to 10k users simultaneously and it would cost me about $1 per megatoken?

I'm simply saying this: there are third party hosters of Open Weight models like deepseek and they have been doing this for a while. Obviously they are not subsidised, do you disagree? If you agree, they have a way to price it at a point that people wanna pay for it and also they aren't losing money. So there's nothing inherent about inference that makes it too costly or whatever.

> I'm simply saying this: there are third party hosters of Open Weight models like deepseek and they have been doing this for a while.

> Obviously they are not subsidised, do you disagree? If you agree, they have a way to price it at a point that people wanna pay for it and also they aren't losing money.

> So there's nothing inherent about inference that makes it too costly or whatever.

Do we have audited GAAP financial data for any of these companies? If we don't, all these are... vibes, man.

Re: AI is slowing down

#662

Post by tempfile »

vessenes wrote: post #590
avarun wrote: post #578

Earlier quoted context omitted.

There is no substance to the article to tackle. Ed Zitron is a clown that makes seven million dollars a year writing unsubstantiated bullshit on the internet. He has been consistently wrong in every single one of his predictions and will continue predicting the sky is falling, because his content has zero educational value. It is purely a confirmation-bias inducing dopamine hit for the perpetually aggrieved.

I like how he's started preliminarily saying "Now listen, I don't make any trades on this fine thinking. And that's okay by me." So far below the standards of public markets due diligence. Sadly credulous readers will follow his advice. Which could turn out to be accurate in the public markets -- who knows? But if it's accurate it will be an accident, because the quality of analysis is so poor we should not call it a…

Not sure what you're getting at here. Would you be happier if he was trading on the arguments he's making, introducing an incentive to lie?

Re: AI is slowing down

#663

Post by dgellow »

jorisw wrote: post #645
dgellow wrote: post #612

Earlier quoted context omitted.

Can we stop spreading the obvious bullshit that is space compute?

Care to at least refer to some sources why?

That has been covered to death since months, there is no good solution to cooling without an atmosphere. But even going into the technical details is a waste of time, think about how complicated maintaining such a system in space would be compared to having it on earth, the whole idea is a complete grift from beginning to end. There is literally no benefit to do it in space other than as a marketing tool.

Nobody who suggests the idea has ever presented a model that is even remotely close to reality

Re: AI is slowing down

#664

Post by baq »

meta_gunslinger wrote: post #625
baq wrote: post #556

Earlier quoted context omitted.

I see plenty if I care to look outside my usual echo chamber. There’s lots of ‘it’s just the next token prediction’ (in 2026 still!), but there are more sophisticated arguments like ‘it can’t be creative’, ‘it doesn’t think’, ‘it’s just pattern matching’ etc. They might even be true for today’s models, but linearly extrapolating an exponential trend is a classic mistake.

Confusing a logarithmic trend with an exponential one is also a classic mistake.

accurate; your point being...?

Re: AI is slowing down

#665

Post by emil-lp »

simianwords wrote: post #654
emil-lp wrote: post #643

Earlier quoted context omitted.

I didn't find the answer there, that's why I asked. What hardware is needed, how much of it, cooling, and what does it all cost you? Or are you saying I can take my old desktop and serve Deepseek v3.2 to 10k users simultaneously and it would cost me about $1 per megatoken?

I'm simply saying this: there are third party hosters of Open Weight models like deepseek and they have been doing this for a while. Obviously they are not subsidised, do you disagree? If you agree, they have a way to price it at a point that people wanna pay for it and also they aren't losing money. So there's nothing inherent about inference that makes it too costly or whatever.

Since you couldn't answer, I asked ChatGPT.

It said: upfront investment: $3M to $6M.

Customers should pay $25k per month.

Checks out

Re: AI is slowing down

#666

Post by oudlys »

fl4regun wrote: post #300
oudlys wrote: post #291

Earlier quoted context omitted.

Sure. But that's not, in my view, how most people use the word productivity when describing LLM use. In my field - operations - productivity is usually described as some rate of production for a specific asset. 100 widgets / machine / hour - for example. "My productivity is 3 PRs / day with the LLM as opposed to 1 PR per every three days". That's how I think people are thinking about it. My point is that's not the sa…

You're correct, I just wanted to add that there is another definition that you may see used online, and it is very specific, and it's important to be aware it's NOT exactly the same thing most normal people mean when they say "productivity".

I appreciate you looking out. Thank you.

Re: AI is slowing down

#667

Post by oudlys »

jurgenburgen wrote: post #536
oudlys wrote: post #291

Earlier quoted context omitted.

Sure. But that's not, in my view, how most people use the word productivity when describing LLM use. In my field - operations - productivity is usually described as some rate of production for a specific asset. 100 widgets / machine / hour - for example. "My productivity is 3 PRs / day with the LLM as opposed to 1 PR per every three days". That's how I think people are thinking about it. My point is that's not the sa…

I’ve noticed more gold-plating. “This random part of the code is slow, I used an LLM to generate a PR that speeds it up.” Okay, you optimized the part that’s not a bottleneck, sped up nothing and cost the company $100 in tokens. Good job?

I'm quite fond of this play on "if a tree falls in the woods":

"If an LLM builds a feature, and no one uses it, did it make value?"

Re: AI is slowing down

#668

Post by pjc50 »

madaxe_again wrote: post #656
pjc50 wrote: post #638

Earlier quoted context omitted.

> Quantitative easing Why though? This isn't a bank run situation. Sure, it's unwise to bet against "US President does something stupid", but the rationale is missing. The oil situation remains very concerning, and you can't print oil.

Oh but it is. The debt that’s funding the AI buildout ends up in the oddest of places - like pension funds. If defaults occur and work their way up the chain, it’s not dissimilar to the subprime affair - and at that point it’s bail out the poor pensioners or get crucified at the ballot box. There’s lots of precedent for this over the past decades, both in the U.S. and over the pond.

How much of this is debt as opposed to equity, though? Can we have a balance sheet?

Re: AI is slowing down

#669

Post by pjc50 »

tipiirai wrote: post #627
munksbeer wrote: post #614

Earlier quoted context omitted.

The reason is obviously that engagement drives his revenue, and correctness or facts have nothing to do with it. Everyone knows that content creators will maximise engagement, and clearly he has found an audience who are seeking out a certain narrative, and will write to that narrative to generate revenue.

"clearly he has found an audience" What do you mean "clearly"? Do you know his engagement numbers? Or you just find the content engaging?

Youtube view/subscribe/like numbers are public. His are .. definitely in the area one can make a living off, but still below popular game streamers.

Re: AI is slowing down

#670

Post by senordevnyc »

dsign wrote: post #154

The way I see it, AI is going to change the world radically. It could be for the worse, the better, or a mix of both, but in my mind there's no doubt. We are only five or six years into the leap LLMs represent. For reference, radio waves were discovered in 1886, Marconi used them for communications in 1895, and while telephone and radio coexisted for many decades, it wasn't until the 1995 that mobile phones and wirel…

I count the start of this AI era as when ChatGPT was released, so we’re actually only 3.5 years in.

And we’re only about a year into the agentic coding era portion of that era.

Post reply on HN