On Intelligence

[ydoaPs]

A while back, bascule recommended to me On Intelligence by Jeff Hawkins. I finally started reading it two days ago and it was great. I would have been done sooner were it not for work

 

Anyway, it gives a pretty detailed view of the neocortex and how it works hierarchically to produce/use invariant auto-associative memories. In doing so, he also tears down the skyhook that many people have that intelligence is some sort of magical thing and replaces it with biology.

 

He argues against artificial intelligence and insists that we should instead create machines with real intelligence using the neocortex as a guide.

 

Overall, I think it's a pretty good book; It's also a really fast read.

 

Also, of interest to the more sciency people here, the book includes an appendix of testable predictions.

[iNow]

Wow... "auto-associative memory." That's a new one for me. Time for more reading.

[bascule]

Jeff Hawkins has a mixed reception among the neuroscience community... seems like a "love him or hate him" sort of thing.

 

That said, for those of you who are too lazy to read, here's a video series with him you can watch:

 

[ydoaPs]

Jeff Hawkins has a mixed reception among the neuroscience community... seems like a "love him or hate him" sort of thing.

 

That said, for those of you who are too lazy to read, here's a video series with him you can watch:

 

 

Since you recommended the book, I'm going to assume you're on the "love him" side. Why is this?

---

Merged post follows:

Consecutive posts merged

One of the interesting aspects of the book, imo, is how it describes that the brain is fundamentally different from a Turing machine.

[bascule]

Since you recommended the book, I'm going to assume you're on the "love him" side. Why is this?

 

I pursue neuroscience for the same direction as Hawkins: I want to know how brains work so I can build them inside of computers. Hawkins theories are quite convenient from my perspective, but as I understand it they're built on theories in neuroscience (namely about the role of the neocortical column) which are still disputed.

 

Here's a video of Jeff Hawkins presenting Hierarchical Temporal Memory to a combined panel of neuroscientists and computer scientists. Skip to 40 minutes in or so to see some realtime peer review:

 

http://video.google.com/videoplay?docid=-2500845581503718756&ei=-WyTSYKsLae4qAPziJHXDA

 

One of the interesting aspects of the book, imo, is how it describes that the brain is fundamentally different from a Turing machine.

 

Yes, if nothing else a Turing machine is sequential whereas a brain is massively concurrent and asynchronous.

[ydoaPs]

I pursue neuroscience for the same direction as Hawkins: I want to know how brains work so I can build them inside of computers. Hawkins theories are quite convenient from my perspective, but as I understand it they're built on theories in neuroscience (namely about the role of the neocortical column) which are still disputed.

 

Here's a video of Jeff Hawkins presenting Hierarchical Temporal Memory to a combined panel of neuroscientists and computer scientists. Skip to 40 minutes in or so to see some realtime peer review:

 

http://video.google.com/videoplay?docid=-2500845581503718756&ei=-WyTSYKsLae4qAPziJHXDA

 

 

 

Yes, if nothing else a Turing machine is sequential whereas a brain is massively concurrent and asynchronous.

 

Sweet link. I'm watching it now.