This is an essential point people seem to misrepresent.

@will_from_mars_ Not sure one can assume data from the optic nerves is quantifiable as bits/bytes. Seems highly speculative and dissonant from reality.

@0xRattata @will_from_mars_ How?

omg totally makes sense. 4-year-olds are SUPER powerful, I bet they'll be a crucial part of the next wave of scientific breakthroughs. they've seen SO much data w/ their visual cortex we could even put them in the enterprise to streamline / overhaul crucial processes!

This is important as hell. There’s definitely a lot more going on in understanding than simply in words.

Why does Lex always look perennially miserable?

I imagine processing 2.5B books would be many times more valuable than 4 years of mostly redundant visual data.

how do u think the books are processed by ur brain

Isn’t most of that data pretty useless though? I ‘learn’ more reading a book than staring at a wall. How much of that data actually gets stored or is used for training? It would be interesting to try and figure out, but I’m not sure how meaningful it would be.

@ylecun The visual information transmitted by our eyes is extremely redundant. Not much to learn from seeing 30-60 frames per second of the same bedroom for six years.

I don't really think the scale of sensory data is why we learn so well. Blind people can learn, even deaf-blind people can learn language well. There's still lots of info from touch and more importantly imo proprioception, but I still don't think it's really scale of sense data that matters. I think we can extract a lot more structure and simply learn more from the data than a llm. He is right about the importance of sensory data though, I just disagree with the framing of scale being why.