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

The changing goalposts of AGI and timelines

mlumiste.com

Post reply on HN
Print view View original

371–380 of 411 posts

Re: The changing goalposts of AGI and timelines

#371

Post by Wowfunhappy »

VorpalWay wrote: post #365
Wowfunhappy wrote: post #336

Earlier quoted context omitted.

Wait, a decision tree is machine learning?

Yes: you fit a decision tree to your dataset in an automated fashion, that fits the definition of machine learning. Just as you would use backpropagation to fit a neural network to your data. This is what I learnt at university some decades ago, and it matches what wikipedia says today: https://en.wikipedia.org/wiki/Decision_tree_learning

Oh, if the tree is made by the computer based on training data, that feels to me like what most people would agree is “artificial intelligence” in 2026 (which is why I think people should actually say “machine learning”).

Re: The changing goalposts of AGI and timelines

#372

Post by parliament32 »

chrysoprace wrote: post #250
djoldman wrote: post #116

Anytime I see "Artificial General Intelligence," "AGI," "ASI," etc., I mentally replace it with "something no one has defined meaningfully." Or the long version: "something about which no conclusions can be drawn because the proposed definitions lack sufficient precision and completeness." Or the short versions: "Skippetyboop," "plipnikop," and "zingybang."

I've largely avoided using the term "AI" to refer to the current LLM and generative technology because it's loaded with too much ambiguity and glosses over the problems with those technologies in the context of conversations around it.

"AI" implies intelligence, which is nowhere to be found. "Text generators" is the best descriptive term.

Re: The changing goalposts of AGI and timelines

#373

Post by Jensson »

computably wrote: post #286
nomel wrote: post #245

Earlier quoted context omitted.

Sorry, I don't understand the question, or how it relates. Are you asking for the current understanding of what specific parts of human intelligence are economically valuable?

I'm pretty sure they were asking for a pinned date for definitions of "economically valuable" and "most (of total economic value)", specifically because, as previous comments noted, the definition and quantity of "economic value" vary over time. If AI hype is to be believed, and if we assume AGI has a slow takeoff, the economy will look very different in 2030, significantly shifting the goalposts for AGI relative to…

Well if humans can do economically valuable mental work the AI can't then its not AGI, don't you think? An AGI could learn that new job too and replace the human, so as long as we still have economically valuable mental work that only humans can do then we haven't reached AGI.

Re: The changing goalposts of AGI and timelines

#374

Post by VorpalWay »

Wowfunhappy wrote: post #371
VorpalWay wrote: post #365

Earlier quoted context omitted.

Yes: you fit a decision tree to your dataset in an automated fashion, that fits the definition of machine learning. Just as you would use backpropagation to fit a neural network to your data. This is what I learnt at university some decades ago, and it matches what wikipedia says today: https://en.wikipedia.org/wiki/Decision_tree_learning

Oh, if the tree is made by the computer based on training data, that feels to me like what most people would agree is “artificial intelligence” in 2026 (which is why I think people should actually say “machine learning”).

That is how decision trees are usually made in my experience. Though I guess you could make one by hand. You could also make a (small) neural network by hand.

In which case you could argue that neither DTs nor NNs are ML. Only the training itself is ML/AI. An interesting perspective, but this will probably just confuse the discussion even further.

Re: The changing goalposts of AGI and timelines

#375

Post by froggit »

godelski wrote: post #301
daxfohl wrote: post #278

Earlier quoted context omitted.

No, independently of OpenAI's definition. If we have AGI there's no reason we'd need to have humans working jobs that only involve typing stuff into a computer and going to meetings all day*. And if all those jobs are eliminated, I guess we'll have bigger problems than to debate whether we've achieved AGI or not. * Which is a much larger class of jobs than just engineering. And also excludes field engineers and other…

> there's no reason we'd need to have humans working jobs that only involve typing stuff into a computer and going to meetings all day I'm not sure I understand, and want to check. That really applies to a lot of jobs. That's all admins, accountants, programmers, probably includes lawyers, and probably includes all C-suite execs. It's harder for me to think of jobs that don't fit under this umbrella. I can think of s…

> [0] farmer, construction worker, plumber, machinist, welder, teacher, doctors, etc

The reason AGI couldn't do these is the lack of a suitable interface to the physical world. It would take a trivial amount of effort for these to be designed and built by the AGI. Humans could be cut from the loop after an initial production run made up of just the subset of these physical interface devices needed to build more advanced ones.

Re: The changing goalposts of AGI and timelines

#376

Post by pu_pe »

sulam wrote: post #367
pu_pe wrote: post #344

Earlier quoted context omitted.

You didn't actually give an example of what the issue with next token prediction is. You just mentioned current constraints (ie generalization and learning are difficult, needs mountains of data to train, can't play chess very well) that are not fundamental problems. You can trivially train a transformer to play chess above the level any human can play at, and they would still be doing "next token prediction". I woul…

I use the chess example because it’s especially instructive. It would NOT be trivial to train an LLM to play chess, next token prediction breaks down when you have so many positions to remember and you can’t adequately assign value to intermediate positions. Chess bots work by being trained on how to assign value to a position, something fundamentally different than what an LLM is doing. A simpler example — without t…

There are chess engines based on transformers, even DeepMind released one [1]. It achieved ~2900 Elo. It does have peculiarities for example in the endgame that are likely derived from its architecture, though I think it definitely qualifies as an example of the fact that simply because something is a next token predictor doesn't mean it cannot perform tasks that require intelligence and planning.

The r in strawberry is more of a fundamental limitation of our tokenization procedures, not the transformer architecture. We could easily train a LLM with byte-size tokens that would nail those problems. It can also be easily fixed with harnessing (ie for this class of problems, write a script rather than solve it yourself). I mean, we do this all the time ourselves, even mathematicians and physicists will run to a calculator for all kinds of problems they could in principle solve in their heads.

[1] https://arxiv.org/abs/2402.04494

Re: The changing goalposts of AGI and timelines

#377

Post by pu_pe »

Otterly99 wrote: post #361
pu_pe wrote: post #344

Earlier quoted context omitted.

You didn't actually give an example of what the issue with next token prediction is. You just mentioned current constraints (ie generalization and learning are difficult, needs mountains of data to train, can't play chess very well) that are not fundamental problems. You can trivially train a transformer to play chess above the level any human can play at, and they would still be doing "next token prediction". I woul…

But chess models aren't trained the same way LLMs are trained. If I am not mistaken, they are trained directly from chess moves using pure reinforcement learning, and it's definitely not trivial as for instance AlphaZero took 64 TPUs to train.

I meant trivial in the sense it's a solved problem, I'm sure it still costs a non-negligible amount of money to train it. See for example the chess transformer built by DeepMind a couple of years ago which I referred to in a sibling comment [1].

[1] https://arxiv.org/abs/2402.04494

Re: The changing goalposts of AGI and timelines

#378

Post by datsci_est_2015 »

bonesss wrote: post #369
datsci_est_2015 wrote: post #354

Earlier quoted context omitted.

“Judgment” is close to my mental model, but I prefer “liability”. All meatspace-based employment comes down to liability. A McDonald’s shift worker is liable for any mistakes made during their shift that are considered their responsibility. A SWE is liable for any mistakes made during their employment that are considered their responsibility. Accountant. Paralegal. CEO. AI will need to be able to experience consequen…

Grok “apologizing” for generating potentially illegal images and related cases working through various legal systems is an early test of that lack of liability shield. There are a number of industries where I don’t think “kinda…” is an acceptable answer to “was this code read before deploying?”. Humans aren’t great at repeating boring tasks ad nauseam.

> Grok “apologizing” for generating potentially illegal images and related cases working through various legal systems is an early test of that lack of liability shield.

You bring up a good point, I forgot that AI will be owned by rich and well-connected people, not the humble masses. If you or I have a small business that uses a delivery AI, we’re liable for murder. If one of the technofascists has a business that uses a delivery AI…

Re: The changing goalposts of AGI and timelines

#380

Post by krackers »

Wowfunhappy wrote: post #336
VorpalWay wrote: post #274

Earlier quoted context omitted.

Machine learning: that definitely includes SVM and regression models. Oh and decision trees. Probably a few other things I'm not thinking of right now. Many people will unfortunately be thinking of just neural networks though. (By the way, if something like a regression model or decision tree can solve your problem, you should prefer those. Much cheaper to train and to run inference with those than with neural networ…

Wait, a decision tree is machine learning?

Expert systems are basically decision trees which are "gofai" (good old fashioned ai) as opposed to deep learning. I've never really seen a good definition for what counts as "gofai" (is all statistical learning/regression gofai? What about regression done via gradient descent?). There's some talk in [1]

[1] https://www.beren.io/2023-04-10-Why-GOFAI-failed/

Post reply on HN