| An expanded Web version of segments seen on CNN |
|
|
Computer mimics human brain at Volvo plant
February 26, 1998
Web posted at: 11:06 a.m. EST (1606 GMT)
DUBLIN, Virginia (CNN) -- What if it were your job to run the
million-square-foot Volvo truck plant in Dublin, while juggling the schedules of 2,300 employees and managing dozens
of other tasks?
Because it is so tough to run a complicated assembly line,
more companies are relying on computers with artificial
intelligence to handle the details.
This intelligence helps a computer find the best solution for
every problem with the help of a so-called genetic algorithm
program that mimics the human brain's decision-making
process.
The father of genetic algorithms, who helped develop the
system, is John Holland, professor of electrical engineering and computers at the University of Michigan.
He told CNN he took his cue from nature: The idea is to "try
to use a computer to breed solutions to problems, much as you
would breed horses ... to get better horses."
The genetic algorithm program lets a computer make decisions
more efficiently than humans.
And that skill is welcomed at the Volvo plant, where the
computer has to figure out how to make 10 different types of
trucks on the same assembly line, each with customized
equipment, while using resources wisely and ensuring the
highest quality of the finished product.
Like playing gin rummy
"In a way, it's like playing gin rummy," said Jamie
Sypniewski of Volvo Trucks North America.
"You only have seven cards. Well, imagine if a human being
had 700 cards in their hand. That's the difference between
using a computer system that can handle a lot more
information and keep track of it -- versus a human being that
doesn't have quite that ability."
To tackle the problem of the assembly line, the computer
first makes small calculations, such as how many dashboards a
two-person crew can install during one shift.
Then the computer takes its best calculations -- its best
genes, one might say -- and applies them to bigger decisions,
such as how many trucks will roll off the line in one day.
"It's best used in areas where you don't really have a good
idea what the solution might be. And it often surprises you
with what you come up with," Holland said.
According to William Calvin, an expert on human intelligence,
the computer's decision-making process is reminiscent of a
Darwinian process, or a process of natural selection.
Calvin explained that the human brain follows a Darwinian
process of natural selection by coming up with many ideas and
rejecting those it deems unfit.
This Darwinian selection of ideas is, of course, not always
used.
"You don't need that sort of thing to do a lot of things in
life. You don't need it to do traditional learning tasks or
to learn new words. But you sure need it every time you look
in the fridge and figure out what to do with the leftovers,"
Calvin said.
No computer really thinks
By that measure, experts say, even the smartest of today's
computers are pretty dumb.
Holland and several other critics say that is true for Deep
Blue, the IBM mainframe computer that defeated world chess
champion Gary Kasparov. Deep Blue actually is quite shallow, they say -- just a big number cruncher with nothing
approaching intelligence, artificial or otherwise.
"The machine, the program, explores all the options, all of
them exhaustively, without any insight, and then picks the
one that's best in that investigation," Holland said.
(
119K/10 sec. AIFF or WAV sound)
Calvin agrees that computers have yet to demonstrate true
artificial intelligence.
"I think that when computers manage to make social
interaction with humans, they would be like a super pet. That
would be one thing that would be very exciting," Calvin said.
(
128K/10 sec. AIFF or WAV sound)
And even though the genetic algorithm computer may seem
intelligent, its inventor admits that it doesn't really
think.
Holland estimated that it might take another 30 years to
invent a computer that really thinks. But, he said, someone
eventually will invent such computers -- and they will
probably be smarter than humans.
Correspondent Rick Lockridge contributed to this report.