AI's no wiz for Shaw Prize winner

Wallis Wang

Artificial intelligence will never replace human beings in their ability to come up with brand new theories, Chinese-American mathematician and this year's Shaw Prize laureate in mathematical sciences Yau Shing-tung said.

Yau is sharing the US$1.2 million (HK$9.36 million) prize with Vladimir Drinfeld, Harry Pratt Judson Distinguished Service Professor of Mathematics at the University of Chicago, for contributions related to mathematical physics, arithmetic geometry and differential geometry.

Almost everything in AI was developed by mathematicians, Yau said, adding it is useful in running math experiments but can make mistakes.

"Even with the famous AlphaGo, sometimes the computer did not work out, maybe one out of 100,000 times it was not working out. This happens because we do not actually understand the mechanism completely so there are still possibilities for making mistakes," he said.

"If you are going to produce some goods, it's probably fine. If you are going to produce 10,000 machines and one or two are not working, that's fine. But if you come to the question of a [fully-autonomous] car, you need to make it 100 percent [safe] There is still a long way to go."

Yau said AI had been important in proving theories, it could also give scientists new ideas and directions but are not fundamental enough.

"Eventually and hopefully, AI could help us figure out some theories, but that's far away."

Still, AI would never be able to replace humans in coming up with brand new concepts, Yau said.

For example, it could never create the concept of quantum mechanics as humans did, as well as Albert Einstein's general relativity, he said.

"This is such a new concept even humans have not understood it completely so I don't believe AI can come up with this radical new concept," Yau said.

AI cannot create something that has never occurred in our history before, no matter how good its database is, he said.

"Human beings are able to do it, but not AI, because AI is all based on the data and all the concepts that have been created before," Yau said.

"Another thing is humans make discoveries based on mistakes as well, we base on things that are mistaken, and AI will not be able to do that."

Yau also said this year's Shaw Prize first recognized his contributions in proving black holes existed using theories for the first time in 1983.

"I was touched when Shaw Prize mentioned my work in proving the existence of black holes as others have never done that before," he said.

Yau said scientists, including himself, have to go through a long and even painful research process but are delighted when a study is done.

"It's like when my wife gave birth to our first child. The delivery took over 20 hours, but I was very happy when my son finally came," he said. "Also, similar to raising a child, we would want to further develop our work and it could lead to cooperation with other subjects."