Artificial Intelligence - Is it a fraud and a Sham?
Well, it's been many years now and the promise
of artificial intelligence ( AI ) still seems to be about where
the promise of controlled fusion is currently - nowhere!
All the talk, both past and present, about the benefits
humans would reap from this pursuit still seems elusive
as ever. I remember, back in the 60's, while working in
a lab, how excited me and my colleagues were when
we got our first computer - it was an IBM7094. One
member of our group "discovered" a program that
would "learn" how to play tic-tac-toe. This was exciting
stuff - a computer would actually "learn" to do something.
Sure enough, the program would lose the first several games
against humans but eventually it "learned" to play at a level
such that it would consistantly play to a draw. Upon analysing
the program, it turned out that all the program was doing
was storing a history of moves and categorized the moves
that would lead to a loss. Eventually, it had enough games
stored so it would no longer lose. What a disappointment!
This was not learning as one would intuit learning to be.
As time progressed, however, it appeared that if computers
became more powerful - the dream of AI would become
a reality. At this time, I left the lab and became a programmer
and the talk of AI still remained in the background with many
people embracing the dream, ie, other programmers and bosses
where I was working believed it would just be a matter of time
before the big "breakthrough" would take place. I remember
a TV show about AI - in the 80's where an AI researcher explained
how his program worked. Somehow he fed some data into the
program about word meanings and word connections,
etc., etc. Then the next day he had the program "write" a story
about an imaginary encounter between a fox and a crow. The
parameters were set and the program "finished" the story and
to the researcher's amazement the story ended with an
unexpected ending. He was estatic, and all I could think of was
bulls__t! This is not AI - because if I had access to the code
and the input data - I could have told him what the ending
would be - it's deterministic - no mystical crap, no deep meaning,
no Einstein thought experiment, just working through the code.
Period!
Computers are not like the human brain - it executes
instructions one at a time in a completely deterministic way. It
processes information serially. Ah, but the human mind, somehow
grabs information from various parts of the brain and in my
opinion, assembles the bits in parallel and like an internal hologram
is processed to give us a thought or concept. This is a process
that allows an Einstein to formulate thought experiments. One
cannot examine the internal connections and the trillions(?) of
possible permutations and "know" how Einstein grasped his
theory of relativity. So, AI cannot involve the storing of data
and the relationships among the data. as this would just lead
to a static state. The human mind is constantly rewiring its neural
network and uses hidden processes to assemble it all to produce
a thought. Computers are static in a sense, and the brain is not.
One processes data serially and the other ( god knows) processes
information in parallel and recognizing patterns in a way that
could be characterized as "all-at-onceness". I think AI research
has its merits - ie, systems that can process huge decision trees
( ie, if this then that, etc.....) like in a game of chess. But for now
I think that a computer capable of human thought is a long way
off. But if a marriage of computers and a living system could be
developed then it may be possible for it ( a cyborg ? ) to someday
derive a Theory of Everything. Much to the chagrin of theoretical
physicists.
|