Haven't read the paper yet, but it's clear that the brain is much more sample efficient than artificial neural networks. We should be spending more time on figuring out why than scaling up LLMs.