"In principle, there's no reason why a computer couldn't be fairly indistinguishable from a human," says Eric Chown, a computer science professor at Bowdoin College who specializes in artificial intelligence. "But that day is a long way off."

Scientists have been working on, dreaming about and arguing over artificial intelligence for decades. While many believe in the possibility of creating "smart" machines, they differ on the ways to go about it, the extent to which these machines will approach human intelligence and the consequences of the ability to create an intelligent machine.

"Films like AI are beneficial because of the interest they spawn in computer science," Chown says, "The problem is they may contribute to unrealistic expectations and they tend to portray AI as ultimately malevolent."

It's important for scientists to consider the possibility of malevolent machines, but there's no reason that a smart machine would necessarily turn against people as is often portrayed in films.

Chown is available to comment on artificial intelligence in general and in relation to the new film "AI". He is a leader in the field of computational models of cognitive mapping, a subset of artificial intelligence. A summary of some of his work follows.

Some of those who study artificial intelligence don't believe we need to rely on the way humans think in order to program robots to "think;" others believe that knowing how the human mind works will help us to make smart computers.

Chown's interest in artificial intelligence stemmed from his interest in human learning. He began researching that topic as a graduate student at the University of Michigan. Now he continues to develop theories of the way people think and uses them to program computers to "think" in similar ways.

Recently he has studied the way that the architecture of the brain helps humans navigate their environment. By doing this, he can program a computer in a similar way to help it navigate around obstacles and between rooms and buildings.

This sort of programming means working with the spatial aspect of our thinking, and the more scientists understand it, the more they can use it to learn how we perform more complicated cognitive tasks.

Spatial thinking controls not only how we get around a room without running into something, but also how we work our way through life. In general, we tend to organize our thoughts in a spatial way. Evolution is parsimonious; when something works for one job, it will tend to be used for as many jobs as is reasonable.

The ability that humans have to represent and think about space has proven to be so useful that we naturally use it in many everyday situations that have nothing to do with navigation. This is one reason why, for example, IQ tests use spatial abilities as one of the fundamental measures of intelligence.

Dealing with artificial intelligence also means considering emotions. "The more intelligent the creature, the more emotional the creature seems to be," Chown says.

"Given the vast amounts of information that people deal with on a daily basis they need simple ways to separate what is important from what can be ignored. Emotions are one way that people do this -- we avoid things that bring negative emotions and try to pursue things that bring positive emotions. Since human needs are so complex and diverse, so too are human emotions."

Chown is also researching emotional computers. To program emotions into a computer, he looks at emotions a bit differently than people are accustomed to.

Because information is so important to humans, and because emotions serve as a way to organize information, emotions are driven by factors such as how good information is (how clear or confused we are), and whether good or bad things are happening to us (pleasure and pain).

These and other factors serve as emotional triggers and are useful in interpreting emotions. (For example, joy is the anticipation of pleasure; fear is the anticipation of pain.) A computer can be programmed to recognize certain actions as causing pleasure, such as plugging into a wall to recharge its energy, or causing pain, such as bumping into a wall.

By doing this, you can get a robot to behave in an emotional fashion -- it could become obsessed with plugs, or keeps its distance from walls.

Some would argue that the computer doesn't really "feel," but if the computer interprets the information as an emotion and acts emotionally, there may be no real difference.

"For me, the question isn't whether or not the computer is truly emotional, it is whether or not it is subject to the same emotional processes that people are," Chown says.

A major contribution of AI as a field of research, Chown says, is that it has taught us how amazing human intelligence is. One of the great challenges lies in the realm of visual recognition. A person can recognize a pen, or another person, from many different angles in many different situations. We haven't yet been able to teach a computer to do this.

"Playing chess isn't nearly as hard as recognizing your mother, and that's something AI has taught us," Chown says. "And AI won't approach human intelligence until this problem is solved."

Chown recently won a $300,000 National Science Foundation Faculty Early Career Development (CAREER) Grant in support of his research. He will use part of the grant to purchase eight high-performance robots to be used in both research and teaching.

These robots will be integrated into several courses currently offered at Bowdoin, and will serve as the basis for a new course specifically about robotics.

Related Links
Bowdoin College
SpaceDaily
Search SpaceDaily
Subscribe To SpaceDaily Express

Thanks for being here; We need your help. The SpaceDaily news network continues to grow but revenues have never been harder to maintain. With the rise of Ad Blockers, and Facebook - our traditional revenue sources via quality network advertising continues to decline. And unlike so many other news sites, we don't have a paywall - with those annoying usernames and passwords. Our news coverage takes time and effort to publish 365 days a year. If you find our news sites informative and useful then please consider becoming a regular supporter or for now make a one off contribution.
SpaceDaily Contributor $5 Billed Once credit card or paypal		SpaceDaily Monthly Supporter $5 Billed Monthly paypal only