There are two groups, people who want to understand the neuroscience and people who want to do computing with the kind of intelligence humans show. The curse is people who start out doing both often seem to get dragged into the neuroscience and ignore the computation.
Doesn’t one inform the other? Doesn’t our neuroscience investigation give us clues about what produces efficient computation?
the HTM software model is computation oriented, from people who spent years doing both.
my bet is as soon as you want to add abilities to that model, since everyone’s clueless on how to build an AGI, “people doing both” would mostly turn to biology to understand what’s missing. Which is in fact the whole paradigm here : try to understand intelligence, then build something that looks like it.
However, I don’t think all computational good-will is sucked up to biology research. From the posts I read here, there are still people such as Paul Lamb I believe, and many others, implementing tweaks to - or experimenting with - the HTM framework, even if they’ve been around here for some time.
By the way, wanted to ask, aren’t there any follow-up to that idea on October of last year, from that young man (Daniel IIRC ?) proposing to code something without minicolumns, with seemingly a very thought-out framework for SMI ?
Yes both can be mutually informative. I think it may be that neuroscience is just so complex people get trapped by the siren call.
As I go through the effort to make an AGI I keep running into roadblocks of understanding.
Each time I go to the biology and find that the problem has been solved a certain way in the brain. This bit of understanding gets me walking on to the next roadblock.
Rinse and repeat.