rhinoceros
Archon
Gender: 
Posts: 1318
Reputation: 8.06
Rate rhinoceros
My point is ...
  
|
 |
Making machines think human
« on: 2003-02-07 14:41:30 » |
 |
To Win Chess, Machines Think Human
Chicago Tribune 02/05/2003
http://www.technologyreview.com/offthewire/3001_522003_2.asp
<snip>
Most computer scientists are not interested in building better chess-playing machines. But they are fascinated with the brain's fantastic pattern-recognition capabilities, and that is what they are trying to build into their new programs.
"Computer chess is kind of a sideshow for most AI people; it's kind of fun and interesting, but it's not exactly mainstream," said Ken Forbus, professor of computer science, education and social policy at Northwestern University.
"Tremendous advances in computer technology have been made, but they won't show up in the chess stuff," said Forbus, who develops artificial intelligence systems for the military and schools. "Things like reasoning systems, planning systems, construction of large knowledge bases and construction of flexible natural language systems - that's going to really make a big change in our lives."
The purpose of artificial intelligence is to build a machine that is more compatible with humans, not like the stubborn donkeys that computers often seem to be today.
Can a learning program be written that will eventually become smarter than its creator, like HAL in "2001: A Space Odyssey"? No one knows.
"By building things that are like us in a lot of ways, we can communicate with them, talk to them and trust them," Forbus said. "They adapt to us instead of us adapting to them, which would be a major thing."
<snip>
MIT is at it again, this time with a humanoid robot that can see itself in a mirror. Cog, which resembles the upper torso of a large man, is programmed to learn by trial and error, getting consistently more proficient, for instance, at repeatedly spotting an object and picking it up.
The brainchild of MIT's AI director Rodney Brooks, Cog uses a camera for eyes and electronics to provide motor skills. It is designed to interact and learn from its environment like young children do.
At the University of Chicago, scientists are trying to develop cyborg technology that would imbed computer chips in the brains of humans.
"We do a lot of mathematical modeling about how the nervous system works, how it stores memories and how it computes," said Dr. John Milton, a U. of C. neurologist. "We would like to build little computer devices that would take over a function that someone has lost because of disease or injury."
But the battle of wits between Kasparov and Deep Junior is also showing why a computer can't be more like a man, at least not in the immediate future.
While Junior runs through millions of possible strategies, Kasparov has said he might consider three - the three best.
People rely on intuition, which is they don't have any hard rules, said Christof Koch, who is researching the interface between humans and machines at the California Institute of Technology. "And that's how chess players play," he said. "They do something that they don't really have a justification for; it just feels right. Computers can't develop the same feelings, but they can develop other ways of learning that allows them to perform in similar ways to a human."
|
Author
Logged
