I changed the subject line to emphasize some issues that have been
debated in AI for over half a century: Since digital computers are
so radically different from the human brain, does it make any sense
to imitate or even waste time by studying neural mechanisms? (01)
A typical analogy is to airplanes, which do not flap their wings.
But that analogy is misleading: The major reason for the Wright
brothers' success is that they studied how birds control their
flight by bending the shape of their wings. So they invented
"wing warping" to control the flight of their airplanes. (02)
As a result, the Wright brothers astounded the world by flying
around Paris while the others could only fly in a straight line. (03)
Moral: You can get a lot of insight from analyzing and reverse-
engineering systems that successfully do what you're trying to do.
But you don't have to imitate or simulate every detail. (04)
On 11/22/2013 5:15 PM, Rich Cooper wrote:
> Thanks for the reference to the slides; it's useful to know
> how many brain regions are associated with behavioral properties,
> especially linguistic behaviors. (05)
This is an area where some "inspiration" from neuroscience can
provide guidance. Jerry Fodor, for example, claimed that there
is a "language of thought" in the brain, which is translated to
and from spoken languages. As evidence, many of our thoughts are
a kind of "inner speech" that resembles our spoken language. (06)
But slide 27 from http://www.jfsowa.com/talks/goal2.pdf (and
the slides that precede and follow it) show the way language is
assembled in Broca's area from information scattered around
many areas of the brain. Some implications: (07)
1. Don't expect to find anything that resembles a language of
thought or a unified "logical form" in the brain. (08)
2. The so-called "inner speech" is not the basis for speech.
Instead, it's generated by the same processes that generate
audible speech -- but with inhibition of the final step. (09)
3. Minsky's _Society of Mind_ with its multiple heterogeneous
modules or agents is a more realistic alternative. See
slides 33 to 36 of http://www.jfsowa.com/goal5.pdf (010)
4. These observations don't mean that logic is irrelevant for AI.
But they show that logic is one among many useful methods used
by people and computer systems. See slide 32 of goal5.pdf. (011)
5. It's also instructive to compare slide 32 of goal5.pdf to
slides 37 and 38 of goal2.pdf. None of these diagrams
influenced the others, but the similarities are remarkable. (012)
James Albus, who drew diagrams that I combined in slide 38, was
a computer scientist who designed successful robotics systems
that were inspired by his study of neuroscience. For a summary,
see http://en.wikipedia.org/wiki/James_S._Albus#Work . (013)
John (014)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (015)
|