News for hackers. Stuff that matters.
Re:The Surprising Effectiveness of Test-Time Training for Ab... by naasking on Tuesday November 12, 2024 @11:42AM (#42117090)
> Because it's a set of puzzles on a 2D grid. We don't live on a 2D grid so it's already on the wrong track.

I don't see what this has to do with anything. Intelligence is about learning patterns and generalizing them into algorithmic understanding, where appropriate. The number of dimensions latent in the dataset is ultimately irrelevant. Humans live in a 4D world, or 3D if the holographic principle is true, and we regularly deal with mathematics 27 or more dimensions. LLMs build models with at least hundreds of thousands of dimensions.

Re:The Surprising Effectiveness of Test-Time Training for Ab... by benchmarkist on Tuesday November 12, 2024 @03:36PM (#42119458)
Show me an LLM that is doing any of the things you mentioned and furthermore I'm willing to bet none of that will be possible after ARC is solved either. How much money would you be willing to bet?
Re:The Surprising Effectiveness of Test-Time Training for Ab... by naasking on Tuesday November 12, 2024 @06:22PM (#42121075)
Not sure what's so controversial, it's well known that LLMs can trivially be viewed as operating in higher dimensional space:

https://gcptips.medium.com/a-geometric-perspective-on-large-...

As for generalizing to algorithms, LLMs don't yet do this as well as humans, but they do do it:

https://arxiv.org/abs/2309.02390

Finally, there's no intrinsic reason why an AI that can reliably solve deductive problems like ARC would be limited to two dimensions.

Re:The Surprising Effectiveness of Test-Time Training for Ab... by benchmarkist on Tuesday November 12, 2024 @06:25PM (#42121103)
Then you have no reason to argue with me.
Re:The Surprising Effectiveness of Test-Time Training for Ab... by naasking on Tuesday November 12, 2024 @06:28PM (#42121143)
The only position I took issue with, and still do, is my closing paragraph of my last post. Your argument for why ARC solvers wouldn't generalize doesn't even make sense.
Re:The Surprising Effectiveness of Test-Time Training for Ab... by benchmarkist on Tuesday November 12, 2024 @07:33PM (#42121624)
No point in arguing. If you think it will generalize then there is no reason to convince random people on the internet that ARC-AGI solver will get you closer to AGI.