While I fell short in understanding all of the technical details in the Emergent Analogical Reasoning paper, seems like a big deal.
Good to see the authors’ substantial caveats at the end of the Discussion section. Human-like reasoning abilities, not a demonstration of machine analogical capability through anything resembling human evolutionary processes.
A very late reply, but yes an interesting paper--I remain pretty suspicious about the way in which they perturbed the Raven Matrices in order to make it more amenable to LMs, and it looks like Melanie Mitchell is also pretty suspicious. You might find her post on it interesting! https://aiguide.substack.com/p/response-to-on-analogy-making-in
While I fell short in understanding all of the technical details in the Emergent Analogical Reasoning paper, seems like a big deal.
Good to see the authors’ substantial caveats at the end of the Discussion section. Human-like reasoning abilities, not a demonstration of machine analogical capability through anything resembling human evolutionary processes.
A very late reply, but yes an interesting paper--I remain pretty suspicious about the way in which they perturbed the Raven Matrices in order to make it more amenable to LMs, and it looks like Melanie Mitchell is also pretty suspicious. You might find her post on it interesting! https://aiguide.substack.com/p/response-to-on-analogy-making-in
Whoops, that's the response post - here's the original https://aiguide.substack.com/p/on-analogy-making-in-large-language
Thanks. Very helpful.
Tiktok is not owned by Tencent.. but ByteDance.
Yep, I think there was a typo in the article that I missed that’s now been fixed. Thanks for the catch!