Is there a leaderboard for the no-restriction version of the competition? I want to see how gpt4 does on it.
Yes there is a secondary leaderboard called ARC-AGI-Pub (in beta) with no limitations: https://arcprize.org/leaderboard
ARC Prize – a $1M+ competition towards open AGI progress
271–280 of 351 posts
Re: ARC Prize – a $1M+ competition towards open AGI progress
#272Earlier quoted context omitted.
> Pay no attention to the man behind the curtain. I could say the same for us, pay no attention to the other humans who are behind the curtain. Humans in isolation are dumb, limited, and can get nowhere with understanding the world. Intelligence is mostly nurture over nature, the collective activity of society nurtures intelligence. It's smart because it learns from many diverse experiences and has a common language…
The A in AGI stands for artificial, so a human+LLM system would not qualify as it has a natural, human component. That doesn't mean it's not an interesting topic, or that it won't help humans discover our world better, it's just the wrong label. Remove the human and you'd just have LLMs talking nonsense at each other. It's not surprising that you get an intelligent system when you include natural intelligence.
The key ingredients are not the humans but the feedback they carry to the model. Humans are embodied and can test ideas in the real world, LLMs need some kind of special deployment to achieve that. It just so happens that chat rooms are such a deployment.
For example, AlphaZero started from scratch and only had feedback from the self-play game outcomes, but that was enough to reach superhuman level. It was the feedback that carried insights and taught the model.
You can make a parallel to the scientific method: you have two stages, ideation and validation. Ideation alone is not scientific. Validation is what makes or breaks ideas. LLMs without a validation system are just like scientists without a lab.
We're not that smart, as demonstrated by the large number of ideas that don't pan out, we can churn ideas fast but we learn from their outcomes, we can't predict outcomes from the beginning and skip validation.
Here is an example of LLMs discovering useful ideas by feedback, even when they are completely outside their training distribution:
"Evolution through Large Models" https://arxiv.org/abs/2206.08896
This works because the task proposed by this paper is easy to test, so there is plenty of feedback. But the LLM still needs to apply ingenuity to optimize it, you can't brute force it by evolutionary methods alone.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#273Earlier quoted context omitted.
> Not just that: people learn mathematics mainly by _thinking over and solving problems_, not by memorising solutions to problems. I think it's more accurate to say that they learn math by memorizing a sequence of steps that result in a correct solution, typically by following along with some examples. Hopefully they also remember why each step contributes to the answer as this aids recall and generalization. The pra…
> The practice of solving problems that you describe is to ingrain/memorize those steps so you don't forget how to apply the procedure correctly Perhaps that is how you learned math, but it is nothing like how I learned math. Memorizing steps does not help, I sucked at it. What works for me us understanding the steps and why we used them. Once I understood the process and why it worked, I was able to reason my way th…
> Perhaps that is how you learned math, but it is nothing like how I learned math.
Memorization is literally how you learned arithmetic, multiplication tables and fractions. Everyone starts learning math by memorization, and only later start understanding why certain steps work. Some people don't advance to that point, and those that do become more adept at math.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#274This is super cool. I share Francois' intuition that the presently data-hungry learning paradigm is not only not generalizable but unsustainable: humans do not need 10,000 examples to tell the difference between cats and dogs, and the main reason computers can today is because we have millions of examples. As a result, it may be hard to transfer knowledge to more esoteric domains where data is expensive, rare, and ha…
> humans do not need 10,000 examples to tell the difference between cats and dogs, I swear, not enough people have kids. Now, is it 10k examples? No, but I think it was on the order of hundreds, if not thousands. One thing kids do is they'll ask for confirmation of their guess. You'll be reading a book you've read 50 times before and the kid will stop you, point at a dog in the book, and ask "dog?" And there is a dev…
I have a small kid. When they first saw some jackdaws, the first bird they noticed could fly, they thought it was terribly exciting and immediately learned the word for them, and generalised it to geese, crows, gulls and magpies (plus some less common species I don't know what they're called in english), pointing at them and screaming the equivalent of 'jackda! jackda!'.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#275Earlier quoted context omitted.
> Not just that: people learn mathematics mainly by _thinking over and solving problems_, not by memorising solutions to problems. I think it's more accurate to say that they learn math by memorizing a sequence of steps that result in a correct solution, typically by following along with some examples. Hopefully they also remember why each step contributes to the answer as this aids recall and generalization. The pra…
> The practice of solving problems that you describe is to ingrain/memorize those steps so you don't forget how to apply the procedure correctly Simply memorizing sequences of steps is not how mathematics learning works, otherwise we would not see so much variation in outcomes. Me and Terence Tao on the same exact math training data would not yield two mathematicians of similar skill. While it's true that memorizatio…
> Simply memorizing sequences of steps is not how mathematics learning works, otherwise we would not see so much variation in outcomes
Everyone starts by memorizing how to do basic arithmetic on numbers, their multiplication tables and fractions. Only some then advance to understanding why those operations must work as they do.
> It's worth noting that for composition, key to abstract reasoning, LLMs failed to generalize to out of domain examples on simple synthetic data.
Yes, I acknowledged that when I said "Composition tasks are still challenging". Comparisons and composition are both key to abstract reasoning. Clearly parametric memory and grokking have shown a fairly dramatic improvement in comparative reasoning with only a small tweak.
There is no evidence to suggest that compositional reasoning would not also fall to yet another small tweak. Maybe it will require something more dramatic, but I wouldn't bet on it. This pattern of thinking humans are special does not have a good track record. Therefore, I find the original claim that I was responding to("there is no AGI pathway in the current research direction") completely unpersuasive.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#276I love this, this is super interesting, but my intuition based on looking at a dozen examples is that the problem is hard, but easy enough that if this problem becomes popular, near-human level results will appear in a year or less, and AGI will not be reached. The problem seems to be finding a generic enough transformation description language with the appropriate operators. And then heuristics to find a very short…
I have the exact same impression.
Imo there's no evidence whatsoever that nailing this task will be true AGI - (e.g. able to write novel math proofs, ask insightful questions that nobody has thought of before, self-direct its own learning, read its own source code)
Re: ARC Prize – a $1M+ competition towards open AGI progress
#277Earlier quoted context omitted.
If a human eye works at say 10 fps, then 8 minutes with a cat is about 10k images :-D
I'd say that was more like a single instance, one interaction with a thing.
But in that single interaction, you might have seen the cat from all kinds of different angles, in various poses, doing various things, some of which are particularly not-dog-like.
I vaguely remember hearing that there's even ways to expand training data like that for neural networks, i.e. by presenting the same source image slightly rotated, partially obscured etc.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#278While I agree with the spirit of the competition, a $1M prize seems a little too low considering tens of billions of dollars have already been invested in the race to AGI, and we will see many times that put into the space in the coming years. The impact of AGI will be measured in trillions at minimum. So what you are ultimately rewarding isn't AGI research but fine tuning the newest public LLM release to best meet t…
Yeah, I also immediately had Dr. Evil narrating the prize money amount in my head once I saw it.
AGI will take much more than that to build, and once you have it, if all you can monetize it for is a million dollars, you must be doing something extremely wrong.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#279I love this, this is super interesting, but my intuition based on looking at a dozen examples is that the problem is hard, but easy enough that if this problem becomes popular, near-human level results will appear in a year or less, and AGI will not be reached. The problem seems to be finding a generic enough transformation description language with the appropriate operators. And then heuristics to find a very short…
I have the exact same impression. Imo there's no evidence whatsoever that nailing this task will be true AGI - (e.g. able to write novel math proofs, ask insightful questions that nobody has thought of before, self-direct its own learning, read its own source code)
I'm not sure the goal of this competition, in and of itself, is AGI. They point to current LLMs emerging from transformers, which in turn emerged from a general basket of building blocks from machine-translation research (attention, etc.). It seems like the suggestion is that to get from where we are now to AGI, some fundamental building blocks are missing, and this is an attempt to spur the development of some of those building blocks, but by analogy with LLMs, the goal here is to come up with a new thing like "attention," not a new thing like GPT4.
Re: ARC Prize – a $1M+ competition towards open AGI progress
#280Earlier quoted context omitted.
> The practice of solving problems that you describe is to ingrain/memorize those steps so you don't forget how to apply the procedure correctly Simply memorizing sequences of steps is not how mathematics learning works, otherwise we would not see so much variation in outcomes. Me and Terence Tao on the same exact math training data would not yield two mathematicians of similar skill. While it's true that memorizatio…
> Simply memorizing sequences of steps is not how mathematics learning works, otherwise we would not see so much variation in outcomes Everyone starts by memorizing how to do basic arithmetic on numbers, their multiplication tables and fractions. Only some then advance to understanding why those operations must work as they do. > It's worth noting that for composition, key to abstract reasoning, LLMs failed to genera…
The point is the memorization exercise requires orders of magnitude fewer examples for bootstrapping.