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

LLMs are a 400-year-long confidence trick

tomrenner.com

Post reply on HN
Print view View original

141–150 of 202 posts

Re: LLMs are a 400-year-long confidence trick

#141

Post by bodge5000 »

schnitzelstoat wrote: post #140
bodge5000 wrote: post #134

Earlier quoted context omitted.

So could the company hiring you to do that work fire you and just use Opus instead? If no, then you cannot compare an engineers salary to what Opus costs, because the engineer is needed anyway. > And API costs at all major providers represent selling the model with a good profit margin. Though we don't know for certain, this is likely false. At best, it's looking like break even, but if you look at Anthropic, they ca…

But now instead of spending 10 hours working on that, he can go and work on something else that would otherwise have required another engineer. It's not going to mean they can employ 0 engineers, but maybe they can employ 4 instead of 5 - and a 20% reduction in workforce across the industry is still a massive change.

Thats assuming a near 100% success rate from the agent, meaning it's not something he needs to supervise at all. It also assumes that the agent is able to take on the task completely, meaning he can go do something else which would normally occupy the time of another engineer, rather than simply doing something else within the same task (from the sounds of things, it was helping with debugging, not necessarily actually solving the bug). Finally, and most importantly, the 20% reduction in workforce assumes it can do this consistently well across any task. Saving 10h on one task is very different from saving 10h on every task.

Assuming all the stars align though and all these things come true, a 20% reduction in workforce costs is significant, but again, you have to compare that to the cost of investment, which is reported to be close to a trillion. They'll want to see returns on that investment, and I'm not sure a 20% cut (which, as above, is looking like a best case scenario) in workforce lives up to that.

Re: LLMs are a 400-year-long confidence trick

#142

Post by stavarotti »

Most of what I've been reading on either side of the argument is reductive. It's possible to have a take based on one's perspective and experience but it's impossible (at this time) to generalize things more broadly. I think what most people feel is multi-faceted: efficiency expectations from "leaders", job change inevitability (perceived or real), economic impact (should things not go well), loss of identity (am I a programmer, engineer, manager of things?), and several others. The discussions on the multiplicative effect of LLMs are being framed as a false dichotomy when it's far more complicated and nuanced.

Re: LLMs are a 400-year-long confidence trick

#144

Post by Traubenfuchs »

falloutx wrote: post #102
Traubenfuchs wrote: post #84

Earlier quoted context omitted.

0. Claude, have a look at frontend project A and backend project B. 1. create a skeleton clone of frontend A, named frontend B, which is meant to be the frontend for backend project B, including the oAuth configuration 2. create the kubernetes yaml and deployment.sh, it should be available under b.mydomain.com for frontend B and run it, make sure the deployment worked by checking the page on b.mydomain.com 3. in fron…

Did you need it though? Like most projects I see being done by people with Claude Code are just their personal projects, which they wouldn't have wasted their time on in the past but now they will get pulled into the terminal thinking its only gonna take 20 mins and they end up burning 100s of subscription dollars on it. If there is no other maintainer & the project is all yours, I dont see any harm in doing it.

I don't NEED this, but turning 2-3 hours fiddling with DTOs, kubernetes yaml, dockerfiles, deployment scripts and other busy work into half an hour does "save you the whole evening" (which our current discussion is about!).

With adults usually having no more than 2-3 hours of free time per work day, this allows you to productively program in your free time without fully burning out.

Also, my company pays for claude and does not give a shit what I do with it.

Re: LLMs are a 400-year-long confidence trick

#145

Post by pmontra »

falloutx wrote: post #85
ACCount37 wrote: post #38

Earlier quoted context omitted.

LLMs of today advance in incremental improvements. There is a finite amount of incremental improvements left between the performance of today's LLMs and the limits of human performance. This alone should give you second thoughts on "AI doomerism".

AI doomerism was sold by the AI companies as some sort of "learn it or you'll fall behind". But they didnt think it through, now that AI is widely seen as a bad thing by general public (except programmers who think they can deliver slop faster). Who would be buying $200/month sub when they get laid off, I am not sure the strategy of spreading fear was worth it. I also don't think this tech can ever be profitable. I h…

The employer buys the AI subscription, not the employee. An employee that sends company code to an external AI is somebody looking for troubles.

In the case of contractors, the contractors buy the subscription but they need authorization to give access to the code. That's obvious if the property of the code is of the customer but there might be NDAs even if the contractor owns the code.

Re: LLMs are a 400-year-long confidence trick

#146

Post by Yossarrian22 »

ACCount37 wrote: post #132
latexr wrote: post #125

Earlier quoted context omitted.

> And there could be a teapot in an orbit around the Sun. I think you’re confused. You are the one making the extraordinary claim, the burden of proof is on you. You asserted LLMs have a finite number of steps to go to reach (overcome?) human limits. You don’t know that. It hasn’t happened. You can’t prove it. I, on the other hand, merely pointed out that is not a certainty. Your teapot argument works against you.

The case against "hard wall": every "hard wall" that was predicted so far was bypassed, and the measured performance keeps going up steadily. The case for "hard wall": wishful thinking.

He’s not saying there is a hard wall he’s saying there’s a point where we’ll need new techniques or technologies not just refine the current one. Less of a hard barrier like the speed of light than an innovative one like creating artificial ammonia to make industrial amounts of fertilizer to support increasing crop amounts

Re: LLMs are a 400-year-long confidence trick

#147

Post by glenneroo »

ACCount37 wrote: post #75
latexr wrote: post #51

Earlier quoted context omitted.

That is not necessarily true. That would be like arguing there is a finite number of improvements between the rockets of today and Star Trek ships. To get warp technology you can’t simply improve combustion engines, eventually you need to switch to something else. That could also apply to LLMs, that there would be a hard wall that the current approach can’t breach.

If that's the case, then, what's the wall? The "walls" that stopped AI decades ago stand no more. NLP and CSR were thought to be the "final bosses" of AI by many - until they fell to LLMs. There's no replacement. The closest thing to a "hard wall" LLMs have is probably online learning? And even that isn't really a hard wall. Because LLMs are good at in-context learning, which does many of the same things, and can do…

Hallucinations are IMO a hard wall. They have gotten slightly better over the years but you still get random results that may or may not be true, or rather, are in a range between 0-100% true, depending on which part of the answer you look at.

Re: LLMs are a 400-year-long confidence trick

#148

Post by sotix »

krystofee wrote: post #32

I disagree with the "confidence trick" framing completely. My belief in this tech isn't based on marketing hype or someone telling me it's good – it's based on cold reality of what I'm shipping daily. The productivity gains I'm seeing right now are unprecedented. Even a year ago this wouldn't have been possible, it really feels like an inflection point. I'm seeing legitimate 10x gains because I'm not writing code any…

> The productivity gains I'm seeing right now are unprecedented.

My company just released a year-long productivity chart covering our shift to Claude Code, and overall, developer productivity has plummeted despite the self-reported productivity survey conveying developers felt it had shot through the roof.

Re: LLMs are a 400-year-long confidence trick

#149

Post by WithinReason »

krystofee wrote: post #32

I disagree with the "confidence trick" framing completely. My belief in this tech isn't based on marketing hype or someone telling me it's good – it's based on cold reality of what I'm shipping daily. The productivity gains I'm seeing right now are unprecedented. Even a year ago this wouldn't have been possible, it really feels like an inflection point. I'm seeing legitimate 10x gains because I'm not writing code any…

The monk analogy is perfect

Re: LLMs are a 400-year-long confidence trick

#150

Post by JacoboJacobi »

sotix wrote: post #148
krystofee wrote: post #32

I disagree with the "confidence trick" framing completely. My belief in this tech isn't based on marketing hype or someone telling me it's good – it's based on cold reality of what I'm shipping daily. The productivity gains I'm seeing right now are unprecedented. Even a year ago this wouldn't have been possible, it really feels like an inflection point. I'm seeing legitimate 10x gains because I'm not writing code any…

> The productivity gains I'm seeing right now are unprecedented. My company just released a year-long productivity chart covering our shift to Claude Code, and overall, developer productivity has plummeted despite the self-reported productivity survey conveying developers felt it had shot through the roof.

I'd like to see a neutral productivity measure? Whether you tell me it went way up or way down I tend to be suspicious of productivity measures being neutral to perception changes that effect expectation, non paradoxical, etc.
Post reply on HN