I didn’t know Bill Atkinson before this triangulation episode.
His anecdote about how Steve Jobs recruited him is fantastic (at about 5:00).
But also he has kept up with the neuroscience.
At about 10:50, he starts talking about the neocortex and HTM theory. It is quite refreshing to hear about HTM theory from someone else than Jeff Hawkins. Not that Jeff explains it badly, but few people except Jeff Hawkins can explain it eloquently, and I’m glad that Bill Atkinson does too.
After Bill Atkinson is done explaining HTM theory, the host Leo Laporte asks a good question (at about 23:00): can this HTM Theory gets you to AGI ? I would have asked: What about reasoning. But it’s the same idea. It’s about the limits of perception/HTM/deep learning with regard to general intelligence.
He gives the argument that a big enough cortex gives you a better prediction device. I think that an AGI or oracle is very different in nature and structure from an HTM hierarchy (event a big hierarchy). I think the answer lies somewhere else.
But still, that was an amazing interview.