July 22, 2005

Kai-Fu Lee

According to Google PR

http://www.google.com/press/pressrel/rd_china.html

Company hires Dr. Kai-Fu Lee as Vice President, Engineering and President of Google China.

Now Microsoft sues Kai-Fu over over Google's hiring.

I worked with Kai-Fu at Microsft. I did not know him in personally, but I saw his work and met him in a couple of meetings. I am almost sure he will continue work on integration of speech recognition and search in Google products.

One interesting note "One Man's Role in the Great Search War" by James Altucher,
RealMoney.com Contributor, http://www.thestreet.com

"I'm writing this because Lee had a big effect on my trading and my career" wrote Mr. Altucher. - "In 1989, I was trying to decide what graduate school in computer science to attend. I flew to Pittsburgh to check out Carnegie Mellon, where Lee was a research associate and a newly minted Ph.D.

I wanted to meet him specifically because of a 1988 paper he wrote, "A Pattern Classification Approach to Evaluation Function Learning," which described his successful attempt at creating a world champion Othello program. His program defeated the reigning champion in 1988.

When I met him, he told me he had been getting blocked on his grad-school research in speech recognition, so he took a year off to write the Othello program. Then he considered doing a similar program for the game Go, thought it too difficult, and eventually finished his Ph.D. in speech recognition, fortunately for speech-recognition users everywhere.


What does this have to do with trading? Here's how his Othello program worked. He compiled a database of thousands of winning positions. For each winning position, he would label it with a vector of variables. Among the variables were how many corners you controlled, how many squares in the center, etc.

Then, given a new random position, he used software (similar to his later speech-recognition software) to determine how close the random position (as identified by its vector of position variables) was to a winning vector. At every level, the software would chose the move that would bring it to a position that was most closely matched (statistically) with a vector from a winning position."

I recommend to read this article entirely to understand how innovation works in 21th centuary (the best way to find this article, use query "Kai-Fu Lee" with http://news.google.com).
Posted 20 years, 2 months ago on July 22, 2005
The trackback url for this post is http://blog.ivolva.com/bblog/trackback.php/10/

Comments have now been turned off for this post