Jeff Hawkins’ “On Intelligence”

Hawkins describes the cortex structure rather clearly. He missed some important points I think. The neuron connects with other neurons. A neuron has an axon and many dendrites. The axon is the output of a neuron and the dendrite is the input. When he says that the pyramid cell’s axon extends far the conclusion is that the pyramid cell is affected by nearly cells, but affects cells far away.
Jargon: X projects to Y means there are signals (nerves) from X to Y.

Regarding Mountcastle’s observations on the uniformity of the cortex, I agree this is an important clue. Consider, however a computer running a Java interpreter which is running a word processing program. Is the computer running Java or the word processor? Mountcastle’s observation is evidence for something vaguely like an interpreter. The computer is doing Java no matter what it is doing, just as the brain is doing the suggested “algorithm” no matter what it is doing.

Hawkins thinks that Turing thought that there was nothing but behavior (via the Teletype). (p. 105) Since Turing specified arguments from behavior Hawkins thinks that Turing doubted anything else was in the picture. That is certainly not the idea that I got in 1950 when I heard of the Turing test. I assumed or was told that the test specified behavior because behavior is manifest as the characters that emerge from the teletype. Awareness, or something isomorphic thereto, was likely necessary, but to convince the skeptic you needed to have something beyond pointing to a computer and saying “Look! now it is contemplating.”. Indeed the human tester at the teletype might require his correspondent to describe its thoughts when it was alone. The human testee is thus on the same footing.

I think that Hawkins has contributed to how we think about the brain. His prediction model is more concrete than the preceding “explanation” that the brain remembers only unusual events. At this level of concreteness I can think of no other explanation. If I were funding a project to build bipedal robots I would demand of my engineers an explanation of the phenomenon Hawkins refers to when a person emits a packet of adrenaline when the floor or a step is not where his brain predicted it would be. If this is important for bipedal animals, is it not for bipedal robots?

I became curious about reptile intelligence as I read. This is unscientific but informative. Here are some snake intelligence test designs and results. I know that mice become curious about new things that are added to their haunts. I wonder if snakes are. Here is a smart lizard.

I see that search engines bring people to this page who are searching for “reptile intelligence”. If someone finds a good source I would appreciate a note to “norm (at sign) cap-lore.com”. Thanks.

New Scientist; 2013 Aug 10. p 8

Machines Come to Life

x, x

“Psychologists believe that humans use a simulated internal self to learn.”

This is in line with Hawkin’s insight that we predict our sensory input and notice mainly the discrepancies. Discrepancies lead to tuning the model. Discrepancies are part of mental state.

Mirror neurons were discovered to furnish a map between limbs of conspecifics and one’s own. Suppose that this extends to mental states.

Make a consciousness note:

Year = 2011. I am sitting in my car parked diagonally and reading a book. Someone gets into the car to my left (this is by sound.). I ‘expect’ to hear an engine start. The next thing I hear is a twig breaking as the car backs out, still no engine noise. It is a Prius. This is an unexpected pattern. Hawkins teaches that we predict the near future events and become conscious of them when prediction does not match observation. I think that I did not become conscious of the next car until I heard the twig break. This suggests a bit of multiprogramming (not multi processing perhaps). This might lead to testable results.