עיתונאיות ועיתונאים, הירשמו כאן להודעות לעיתונות שלנו
הירשמו לניוזלטר החודשי שלנו:
שתף
It takes less than a second from the time your eyes register an image until your brain tells you what you are seeing. But what happens during that interval? How does the brain process visual images, and how can it do this so much better than any machine?
Profs. Amiram Grinvald and Ad Aertsen and Dr. Amos Arieli -- all of the Neurobiology Department -- are working with Prof. Shimon Ullman of the Applied Mathematics and Computer Science Department to find the answers by studying how groups of brain cells interact with one another to process visual information. In experiments, the scientists show pictures to subjects and monitor their brain activity by optical imaging, a technique pioneered by Grinvald. They also use theoretical mathematics to develop models of visual object recognition, and to interpret and predict the effects on brain activity of different attributes of the pictures, such as their angle and distance, and whether they are moving or still. The benefits of working together go both ways, according to Grinvald and Ullman.
"The brain is the most successful computing machine known," says Grinvald. "For neurobiologists, mathematics and computer science provide tools that enable us to analyze the results of our experiments and construct theoretical models that offer predictions we can test. For mathematicians who want to create artificial intelligence, the brain is a model that already does almost everything they want to imitate, so they can benefit from understanding how it operates."
Research by Arieli and others in Grinvald's team, reported in Science, has already shown that an image produces different brain activity patterns in the same individual at different times, depending on the viewer's state of mind. Significantly, however, the team has now found that if internal activity is removed, a small "core" portion of brain activity is the same whenever the same image is presented.
Further clarification of these processes is likely to have significant implications for two broad areas: brain research, where it may help explain how the brain accomplishes higher functions, and computers, where scientists hope to create intelligent artificial systems. For example, a computer can scan a picture of a face but when shown that face from a different angle, it does not recognize it. An "intelligent" computer would recognize the face from any angle or in motion, just as the brain does. Brain-like methods that might allow artificial vision systems to recognize objects in a wide range of conditions were described in Ullman's recent book High-Level Vision (MIT Press).
"Our project is a blend of computer science and brain science," says Ullman, "and we believe our work is important for both technology and biology."
שתף
