Hacker News Reader

Comment by aheilbut | Hacker News Reader

Minsky talked about the 'society of mind' involving competition between many autonomous modules a very long time ago; funny that he isn't even mentioned.

The brain-as-computer analogy needs more qualification than it is usually given. When one says that, people tend to think of a computer very abstractly; as some Turing-machine like substrate on which one can run any computation. The brain isn't like that at all; it's built out of all kinds of ASICs that have evolved for very specific functions. The reality is probably much closer to 19th century phrenology than 50s cybernetics.

View on HN

Then how do you explain that:

http://en.wikipedia.org/wiki/Hemispherectomy

One case, demonstrated by Smith & Sugar, 1975; A. Smith 1987, demonstrated that one patient with this procedure had completed college, attended graduate school and scored above average on intelligence tests. Studies have found no significant long-term effects on memory, personality, or humor after the procedure, and minimal changes in cognitive function overall.

The brain has bilateral symmetry. Don't try that trick cutting along a different axis.

Also, they don't talk much about motor function and vision.

More meat for the AI warriors is found on Minsky's homepage: http://web.media.mit.edu/~minsky/

> The brain isn't like that at all; it's built out of all kinds of ASICs that have evolved for very specific functions.

That turns out not to be the case. There are some specialized modules, but stroke studies have shown that massive damage to specific functions can often be recovered by retraining. It seems that particular parts of the brain are especially good at learning particular jobs, and that when they learn a job they somehow suppress redundant learning by other regions. Much of the apparent specialization of the brain is very possibly an optimization process, akin to a computer cluster that automatically assigns I/O jobs to nodes near the mass storage.

I believe that optimizations observed are almost purely function of brain activity. Any boosts genes provide to specific regions are quantitative not qualitative in nature.

In the cerebral cortex, I agree. It seems like any handy chunk of cortex with enough connections can learn a task.

Outside the cortex there are dedicated organs within the brain: circadian rhythym generators, the locus ceruleus, the hippocampus, and many others.

Maybe it'd be more interesting to think of it like an FPGA that can run a Verilog JIT compiler and reprogram itself when damage occurs or new information changes priorities?

Reply on news.ycombinator.com