François CholletVerified account

I'm sorry, I wasn't aware I was only allowed to tweet novel ideas. Of course this is old and well-known. Yet many folks still haven't fully internalized it. Does everyone in your field behave like this (in public, at that) or is it just you guys?

I love the fact that you recognize what the interesting questions are in our field, as I'm sure I couldn't even figure out what separates the interesting from mundane questions in yours. It's exciting any time someone with different expertise shares our awe in what we study.

Cool, and I appreciate your Neisser reference. Active perception is still not a hot topic in AI today. I used to do some research in that area (active vision with an anticipative eye saccade model) in 2012, and back then these ideas had very little traction. It's trending up tho.

That's actually exciting. The fact that active perception is not a hot topic in AI is amazing to me, as it's almost taken for granted in our field. Is that bc its hard to build it into AI, or does it represent a fertile avenue for collaboration between cognitive & AI researchers?

It has simply not yet shown to be necessary, or even useful. It actually seemed like a more attractive avenue when we knew less and our models performed worse.

ML models don't attempt to emulate human cognition, and they're solving a different problem than embodied cognition in the first place, with different constraints and different degrees of freedom.

If your input is a static image that you're trying to classify, that's a very different setup than being an embodied agent immersed in a dynamic world subject to cause and effect. In the former case, processing all the information available in one go is actually more effective