Block · one region of the page, as the scanner read it. It may hold a whole story, part of one, several, or an advertisement; stitching blocks into articles is the next step. Text is supplied OCR.

Page 5 · column 1 of 6 · from the scan, no model involved

The clipping this text was read from
The clipping this text was read from

Worried about frost damage to * his apple trees, the former hurries ; into his house to consult - his com| puter. The computer asks the | former some specific questions, and • suggests various remedies when he • supplies the required information. ! The computer can also advise on 1 pest management and drought con| trot J At her office in town, the doctor j wondering whether to prescribe 2 penicillin to a patient with a viral in! fection, but who is also taking antiI histamines for an allergy problem. ! The doctor is concerned about the * possible dangers of drug interaoj tions. She, too, turns to her comi puter.

These scenes, which may be commonplace in a few years, show “expert systems” in action. Expert systems are computer programs containing the knowledge of human experts in narrowly defined subjects such as apple orchard management, or combination drug therapy. Expert systems consist of two parts — a knowledge base, and information which tells the system how to use that knowledge in a logical, reasoning way to produce solutions to problems.

These smart computers are an example of how artificial intelligence may be commercially applied.

Artificial intelligence refers to the wide range of computer programs that allow robots and other machines to mimic human behavior - to have sensing and movement capabilities, as well as the ability to think rationally, learn from experi; - ence , and make inferences.

Progress jn the three decades-old field of artificial intelligence has been enormous, but uneven. Scientists and engineers have succeeded in making robots sensitive to temperature variations, but have not figured out how to give them a sense of taste, texture, or smell.

Though duplicating the precise and complex action of human hands is difficult, researchers at the Massachusetts Institute of Technology have been able to devise a three-fingered robot hand which can feel the shape and size of an object, “think” about the ways the object can be held securely, and grasp it in these positions.

At Virginia Tech, John Roach is working on multi-armed robots. A lot of robot research in the United States, he says, is focused on onearmed robots, but many tasks can’t be done with only one arm — unscrewing the cap off a jar, or hammering in a nail, for instance. “We’re trying to get two or more arms to work together efficiently and avoid banging each other - easy for humans, but not for robots,” said Mr. Roach, an associate professor of computer science.

Like robotic motion, expert systems have generated considerable excitement in universities and industries because of the technology’s usefulness and market potential.

POMME, the orchard management system to be tested soon by

83.6%