2024-12-09 at 19:20

humans learn not from repeated correct examples but from the edge cases of mistakes (wait we also do learn from repeating correct things like w spaced repetition, right? and also w reviewing the same practice problem ... oh wait but . perhaps that is also still actually traveling the edge case of mistakes, bc otherwise tbh we just skip the example when we're studying)

eg dekeyser's skill building

oh wait i guess this is what RL does (?)

==========

also why havent we created a multimodal model that learns modal associations like how we learn them .? (not just input (audio to experience) but also output (mimicking audio and associating that speech with experience))

would we even be able to simulate that ..? bc the ai lives in a world where the sensory inputs are so much different (mp3 files and png files instead of ears and eyes)

but why havent we attempted it?

ugh maybe im just being stupid and maybe i should learn "the bitter lesson"

and also maybe LLMs are already extremely efficient compared to brains (like 1B-500B params compared to like 100 Trillion in brains) but idk man . i feel like this is promising ! i do feel like at least some of what im saying is promising and things that have not been tried yet

==========

Automated Reasoning Basics | Douglas Lenat and Lex Fridman
https://www.youtube.com/watch?v=-GQPrwPghZY

"""
so suppose you're building one of these applications and the system gets some answer wrong, and you say "oh yeah the answer to this question is this one, not the one you came up with", then what the system can do is it can use everything it already knows about common sense, general knowledge, the domain you've already been telling it about, and context like we talked about, and so on, and say "well here are seven alternatives each of which i believe is plausible given everything i already know and if any of these seven things were true i would have come up with the answer you just gave me instead of the wrong answer i came up with -- is one of these seven things true?" and then you the expert will look at those seven things and say "oh yeah number five is actually true" and so without actually having to tinker down at the level of logical assertions and so on um you'll be able to educate the system in the same way that you would help educate another person who you were trying to apprentice or something like that. so that that significantly reduces the mental effort or significantly increases the efficiency of the teacher the human teacher exactly and it makes more or less anyone able to be a teacher in that way. so that's that's part of the the answer and then the other is that uh the system on its own will be able to um through reading through conversations with other people and so on learn the same way that you or i or other humans do you
"""

bro this is rly cool
id like this to happen
i wonder to what extent this already happens and i wonder how models learn to answer correctly rn, like how does RL + gradient descent actually happen
i wonder how similar/dissimilar this is to RL and gradient descent -- and like . this is a lot like how we learn, and im wondering why cant AI learn like we learn

and this also reminds me of my method of brainstorming the functions for ARC tasks