15 years ago

[fireproof]

Ran into this NY Times article from 2002 this morning:
http://www.nytimes.com/ref/open/thiswee ... -OPEN.html

"I think in the long run the only way to write a strong Go program is to have it learn from its own mistakes, which is classic A.I., and no one knows how to do that yet," Mr. Fotland said.

It's giving me a good perspective on the 7-8 years I've been playing, and how far computer Go has come.

[EdLee]

I think in the long run the only way to write a strong Go program is to have it learn from its own mistakes, which is classic A.I., and no one knows how to do that yet

Funny thing is neural nets and Monte Carlo have been around, independently, for quite some time; but it took some smart folks to try one, and then put them together.
Also, maybe we needed to wait for 2015 hardware and the massive scales of the companies, like IBM and Intel in 1997.
Very good fortune to have Demis Hassabis.

[hyperpape]

As someone who does zero machine learning/AI, I have a hard time telling, but I think there were legitimately new ideas and techniques for working with neural nets that became widely used in computer science sometime around 2010 (what's called deep learning).