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

A bear case: My predictions regarding AI progress

lesswrong.com

Post reply on HN
Print view View original

131–140 of 220 posts

Re: A bear case: My predictions regarding AI progress

#131

Post by Imnimo »

>Test-time compute/RL on LLMs: >It will not meaningfully generalize beyond domains with easy verification.

To me, this is the biggest question mark. If you could get good generalized "thinking" from just training on math/code problems with verifiers, that would be a huge deal. So far, generalization seems to be limited. Is this because of a fundamental limitation, or because the post-training sets are currently too small (or otherwise deficient in some way) to induce good thinking patterns? If the latter, is that fixable?

Re: A bear case: My predictions regarding AI progress

#132

Post by bilsbie »

I have times when I use an LLM and it’s completely brain dead and can’t handle the simplest questions.

Then other times it blows me away. Even figuring out things that can’t possibly have been in its training data.

I think there are groups of people that have either had all of the first experience or all of the latter. And that’s why we see over optimistic and over pessimistic takes (like this one)

I think the reality is current LLM’s are better than he realizes and even if we plateau I really don’t see how we don’t make more breakthroughs in the next few years.

Re: A bear case: My predictions regarding AI progress

#133

Post by aetherson »

orangebread wrote: post #84

I think the author provides an interesting perspective to the AI hype, however, I think he is really downplaying the effectiveness of what you can do with the current models we have. If you've been using LLMs effectively to build agents or AI-driven workflows you understand the true power of what these models can do. So in some ways the author is being a little selective with his confirmation bias. I promise you that…

I don't think he is downplaying the effectiveness of what you can do with the current models. Rather, he's in a milieu (LessWrong), which is laser-focused on "transformative" AI, AGI, and ASI.

Current AI is clearly economically valuable, but if we freeze everything at the capabilities it has today it is also clearly not going to result in mass transformation of the economy from "basically being about humans working" to "humans are irrelevant to the economy." Lots of LW people believe that in the next 2-5 years humans will become irrelevant to the economy. He's arguing against that belief.

Re: A bear case: My predictions regarding AI progress

#134

Post by fragmede »

roncesvalles wrote: post #56
csomar wrote: post #22

> LLMs still seem as terrible at this as they'd been in the GPT-3.5 age. Software agents break down once the codebase becomes complex enough, game-playing agents get stuck in loops out of which they break out only by accident, etc. This has been my observation. I got into Github Copilot as early as it launched back when GPT-3 was the model. By that time (late 2021) copilot can already write tests for my Rust function…

The last line has been my experience as well. I only trust what I've verified firsthand now because the Internet is just so rife with people trying to influence your thoughts in a way that benefits them, over a good faith sharing of the truth. I just recently heard this quote from a clip of Jeff Bezos: "When the data and the anecdotes disagree, the anecdotes are usually right.", and I was like... wow. That quote is t…

> it's just not there

Build the much maligned Todo app with Aider and Claude for yourself. give it one sentence and have it spit out working, if imperfect code. iterate. add a graph for completion or something and watch it pick and find a library without you having to know the details of that library. fine, sure, it's just a Todo app, and it'll never work for a "real" codebase, whatever that means, but holy shit, just how much programming did you need to get down and dirty with to build that "simple" Todo app? Obviously building a Todo app before LLMs was possible, but abstracted out, the fact that it can be generated like that's not a game changer?

Re: A bear case: My predictions regarding AI progress

#135

Post by klik99 »

This almost exactly what I’ve been saying while everyone was saying we’re on the path to AGI in the next couple of years. We’re an innovation / tweak / or paradigm shift away from AGI. His estimate in the 2030s that could happen is possible but optimistic- you can’t time new techniques, you can only time progress on iterative progress.

This is all the standard timeline for new technology - we enter the diminishing returns period, investment slows down a year or so afterwards, layoffs, contraction of industry, but when the hype dies down the real utilitarian part of the cycle begins. We start seeing it get integrated into the use cases it actually fits well with and by five years time its standard practice.

This is a normal process for any useful technology (notably crypto never found sustainable use cases so it’s kind of the exception, it’s in superposition of lingering hype and complete dismissal), so none of this should be a surprise to anyone. It’s funny that I’ve been saying this for so long that I’ve been pegged an AI skeptic, but in a couple of years when everyone is burnt out on AI hype it’ll sound like a positive view. The truth is, hype serves a purpose for new technology, since it kicks off a wide search for every crazy use case, most of which won’t work. But the places where it does work will stick around

Re: A bear case: My predictions regarding AI progress

#136

Post by _Algernon_ »

rchaud wrote: post #115
roncesvalles wrote: post #64

Earlier quoted context omitted.

They were revolutionary as product genres, not necessary individual companies. Ordering a cab without making a phone call was revolutionary. Netflix at least with its initial promise of having all the world's movies and TV was revolutionary, but it didn't live up to that. Spotify because of how cheap and easy it was to have access to all the music, this was the era when people were paying 99c per song on iTunes. I've…

> They were revolutionary as product genres, not necessary individual companies. Even then, they were evolutionary at best. Before Netflix and Spotify, streaming movies and music were already there as a technology, ask anybody with a Megaupload or Sopcast account. What changed was that DMCA acquired political muscle and cross-border reach, wiping out waves of torrent sites and P2P networks. That left a new generation…

>every mobile OS vendor

Maybe half? Android has consistently had this capability since its inception.

Re: A bear case: My predictions regarding AI progress

#137

Post by rahimnathwani »

carlosdp wrote: post #112
stego-tech wrote: post #2

> At some point there might be massive layoffs due to ostensibly competent AI labor coming onto the scene, perhaps because OpenAI will start heavily propagandizing that these mass layoffs must happen. It will be an overreaction/mistake. The companies that act on that will crash and burn, and will be outcompeted by companies that didn't do the stupid. We're already seeing this with tech doing RIFs and not backfilling…

I'll take that bet, easily. There's absolutely no way that we're not going to see a massive reduction in the need for "humans writing code" moving forward, given how good LLMs are getting at writing code. That doesn't mean people won't need devs! I think there's a real case where increased capabilities from LLMs leads to bigger demand for people that know how to direct the tools effectively, of which most would proba…

  increased capabilities from LLMs leads to bigger demand for people that know how to direct the tools effectively
This is the key thing.

Re: A bear case: My predictions regarding AI progress

#138

Post by weweersdfsd »

danielbln wrote: post #69
tcoff91 wrote: post #46

Earlier quoted context omitted.

A midlevel web developer should do a whole lot more than just respond to chat messages and do exactly what they are told to do and no more.

When I use LLMs that what it does. Spawns commands, edits files, runs tests, evaluates outputs, iterates and solutions under my guidance.

The key here is "under your guidance". LLM's are a major productivity boost for many kinds of jobs, but can LLM-based agents be trusted to act fully autonomously for tasks with real world consequence? I think the answer is still no, and will be for a long time. I wouldn't trust LLM to even order my groceries without review, let alone push code into production.

To reach anything close to definition of AGI, LLM agents should be able to independently talk to customers, iteratively develop requirements, produce and test solutions, and push them to production once customers are happy. After that, they should be able to fix any issues arising in production. All this without babysitting / review / guidance from human devs, reliably

Re: A bear case: My predictions regarding AI progress

#139

Post by fragmede »

InfiniteTitan wrote: post #87
csomar wrote: post #67

Earlier quoted context omitted.

> None of those are revolutionary companies. Not only Uber/Grab (or delivery app) were revolutionary, they are still revolutionary. I could live without LLMs and my life will be slightly impacted when coding. If delivery apps are not available, my life is severely degraded. The other day I was sick. I got medicine and dinner with Grab. Delivered to the condo lobby which is as far as I can get. That is revolutionary.

Is it revolutionary to order from a screen rather than calling a restaurant for delivery? I don’t think so.

Practically or functionally? Airbnb was invented by people posting on craigslist message boards, and even existed before the Internet, if you had rich friends with spare apartments. But by packaging it up into an online platform it became a company with 2.5 billion in revenue last year. So you can dismiss ordering from a screen instead of looking at a piece of paper and using the phone as not being revolutionary, because of you squint, they're the same thing, but I can now order take out for restaurants I previously would never have ordered from, and Uber Eats generated $13.7 billion in revenue last year, up from 12.2.

Re: A bear case: My predictions regarding AI progress

#140

Post by fragmede »

hattmall wrote: post #34
csomar wrote: post #25

Earlier quoted context omitted.

Nope. I try the latest models as they come and I have a self-made custom setup (as in a custom lua plugin) in Neovim. What I am not, is selling AI or AI-driven solutions.

Similar experience, I try so hard to make AI useful, and there are some decent spots here and there. Overall though I see the fundamental problem being that people need information. Language isn't strictly information, and the LLMs are very good at language, but they aren't great at information. I think anything more than the novelty of "talking" to the AI is very over hyped. There is some usefulness to be had for su…

what does subsidization have to do with your use of a thing?
Post reply on HN