By GARRY BARKER
TECHNOLOGY REPORTER
THE AGE

Wednesday 26 January 2000

Professor Hugo de Garis, physicist, lately of Melbourne and now of Kyoto in Japan, fears that his experiments may ultimately lead to the extermination of the human race. What do you think?

At the Kyoto Advanced Telecommunications Research Institute, Professor de Garis switched on a machine with which he will build the world's first neural circuits for a true artificial brain.

In the next 12 months the cellular automata machine (CAM) in his laboratory will create a device composed of 75million silicon neurons, similar in capability to those in a human brain.

The neuron networks are built up so that their connections are random, as they are in the human brain. Most of them fail in production and are discarded by a system based on Darwin's theory of evolution. Even so, the circuits are built, tested, accepted or rejected at blinding speed, many thousands every minute.

When it is finished some time in 2001, this artificial brain or "artilect" will go into a four-legged robot called Robokitty.

By then work will have begun on the next generation of the artificial brain which, Professor de Garis says, could be finished about 2007 and would have more than 10 billion neurons. This would bring it to about the level of a village idiot but within reach of the 23billion organic neurons contained in the cortex of a human male (19 billion in a female).

Then comes the third generation, which Professor de Garis expects to be finished about 2011 - a fearsome creation of 1000 billion neurons, vastly larger than that of a human.

"By then," says this unconventional Australian, "I expect we'll be in a debate about whether we should proceed any further. "Long-term I am very worried about the political impact of brain building.

"Since I am helping to pioneer this brain-building field, I feel a strong moral obligation to stimulate discussion on this enormous question.

Do we allow the artificial intellects to take over or not?" Futurologists, such as the American computer engineer and author Ray Kurzweil, agree with him. While they themselves are riding and driving the technological revolution, they also see its scary side.

A massively powerful artificial brain could easily develop contempt for its comparatively puny human makers, says Professor de Garis, who predicts that such a question could be this century's burning issue.

On one side will be those afraid of the consequences of the science. On the other those who see it as part of human destiny and who say that if artilects are created by humans, then humans can set the boundaries for the artificial intelligence.

Professor de Garis is not so sure about humans retaining control, particularly when it comes to a silicon brain 40 times smarter than your average man. These, he says, should be coming out of the CAM machines by the second half of this century.

Some see parallels with the debate raised by the cloning of Dolly the sheep.

The CAM machine with which Professor deGaris is working was built by Genobyte, a US company based in Boulder, Colorado. It produces microscopic modules on silicon chips each of about 1000 artificial neurons. Such electrical connections in our human brains control our movements, our senses and, perhaps most ominously when it is seen in an artificial environment, our emotions and our imaginations.

In his profile on his personal website, the professor says: "My dream in life is to build artificial brains with billions of artificial neurons, and see brain-like computers become a trillion-dollar industry within 20 years."