Could AI solve this puzzle? (ARC-Game)
ฝัง
- เผยแพร่เมื่อ 26 ก.ย. 2024
- #ai #arc #puzzle
We play the ARC-Game.
Game: volotat.github...
Code: github.com/vol...
Original Dataset: github.com/fch...
Links:
TabNine Code Completion (Referral): bit.ly/tabnine-...
TH-cam: / yannickilcher
Twitter: / ykilcher
Discord: / discord
BitChute: www.bitchute.c...
Minds: www.minds.com/...
Parler: parler.com/pro...
LinkedIn: / yannic-kilcher-488534136
BiliBili: space.bilibili...
If you want to support me, the best thing to do is to share out the content :)
If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
SubscribeStar: www.subscribes...
Patreon: / yannickilcher
Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq
Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2
Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m
Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
There is an easier way to fill space with color by dragging it from the palette :) Thanks for such an amazing video. It was a pleasant surprise to see it here and in ML news.
What a pleasure that authors are involved in natural development of their work.
I have not thought about this too much, but I feel like generalized Procrustes could be used as a baseline for solving some problems, esp. where certain problems demonstrate patterns like symmetry since the solvers explicitly solve for that.
What about if one used the image to generate the weights of an MLP that can predict the values at each pixel of the predicted output shape? Seems like it could be an effective solution, but I'm interested in what you guys think
Human inductive biases. All of them.
If this is your reasoning as to why AI solutions today can't perform on this benchmark, it seems to me that predicting the next word in a conversation that LLMs are good at seem to be a better example of human-biased inductive reasoning.
@@jhanschoo If an LLM can't count the number of r in strawberry, then how can it zero shot learn abstract concepts like : inside, moving towards, symmetry, path, behind, etc... ? We are asking a model to infer one-shot concepts we are born with. It's a pretty difficult task. I think it will be solved with sufficient pretraining; but then it won't prove that model is intelligent.
Im subscribin. You need P r o m o S M!!