Question that’s been bothering me. I remember in one of the Numenta videos (don’t remember which otherwise I’d post a link) Jeff mentions the neocortex didn’t evolve like the older parts of the brain, which evolved over millions of years, instead the neocortex rapidly developed with a repeating circuit. I don’t recall him elaborating more and forgive me if I am miss representing what he said as again I can’t find that video. Anyway if the neocortex didn’t evolve like everything else then what explains its development?
1 Like
I’m not sure about the talk you’re thinking of, but I know in On Intelligence Jeff talks a bit about Vernon Mountcastle’s work and its impact on his thinking. He points out the significance that the neocortex essentially has the same basic learning mechanism throughout, rather than a variety of different systems working together. Therefore the process of evolution only needed to duplicate a basic function rather than produce a long series of novel ones.
2 Likes
Yeah of course it took a really long time for the structure of isocortex (what Jeff calls neocortex) to develop out of the precursor structures (paleocortex, archicortex, periallocortex …). I think the thing you (JamesLikesBrains) are remembering is that once a kind of cortex (e.g. isocortex) exists in an animal lineage, it can get bigger really fast. (Well, fast compared to other things—it may still take tens of thousands of years.) The reason it can get bigger fast is because cortex consists of lots of copies of a little unit (cortical columns, or maybe cortical minicolumns, depending on who you ask). You don’t need to change many genes to increase the number of copies of this unit; and increasing the number of copies may well be evolutionarily adaptive, I think for vaguely similar reasons as the fact that deep neural networks give better answers (after training) if you just stack more layers. (Emphasis on “vaguely”, yes I know it’s not a perfect analogy.)
1 Like