Hacker News Reader

Comment by isaacfrond | Hacker News Reader

I think the Go example is facetious. For the longest time, Go simply was not amenable to search. It is only because of breakthrough in theory that Go became amendable to search. And, it is hardly a straightforward search mind you. Two self learning networks are involved. A first to guide the expansion of a Monte Carlo based tree search, a second to evaluate the nodes in the tree. You can hardly blame a researcher in the 70s, 80s, 90s, 00s for not focusing on computer power, when you have 19^2 expansion factor for each move.

View on HN

Go has been described as the drosophila of AI. Once a program will know how to play go, we'll be able to make more general artificial intelligences. This is exactly what has happened with alphago and chatgpt. We're not repeating the same thing that happened with chess, but we are past the singularity. Recent advances in AI are lighting up the very nature of intelligence. Our memory and our brain are a machine for predicting the future. It's very similar to what recent models do. Alphagozero has rediscovered the history of chess openings in the same order as humans. We are no longer discovering simple new algorithms supported by machine advances, but architecting predictive systems based on guided learning. When we developed the algorithm for playing chess, the possibilities for generalization were limited. With these new architectures, new possibilities are appearing every day. IMHO, the predictions were true: since alphago, we've entered a new world.

>> Go has been described as the drosophila of AI.

John McCarthy described chess as "the drosophila of AI":

http://jmc.stanford.edu/articles/drosophila/drosophila.pdf

Who was it that described Go as the drosophila of AI?

It was John McCarthy, in the very paper that you cite!

See section 7:

As a fourth Drosophila I would like to mention the research on Computer Go.

I finished my studies in 1993. Artificial intelligence was my major and I had done a study on programming the game of Go (a review of research state). At that time, the success of brute force for chess had caused much disillusionment in the AI world, which had hoped to see a drosophila. The transfer of this hope to the game of go was present in numerous publications. Even if it doesn't show up very well in a Google search, there are plenty of allusions in articles about programming the game of Go. Since the creation of go language, it is more difficult to find.

Oh, oops. I missed that. I linked the pdf thinking it was a different text.

Thanks for the correction.

There's no real knowledge transfer from Alphago to Chatgpt. They share some underlying techniques, but they sit on different branches of a tree, one is not built on top of the other.

It's not called "the bitter lesson" for being optimistic.

One of the points, I think, is that there is very little a researcher in the 70s could have done to make progress on these problems, because the computing power just wasn't there yet; and once the computing power was here, almost all of that 70s researcher's work became obsolete.

i don't know. it's mysterious that nobody even came up with the algorithms for decades, which often do still work well on toy examples. MCTS is the kind of thing people definitely could have thought of at any point since 1950; and neural network value functions for board games had already succeeded with TD-GAMMON.

this is even more striking in poker, where counterfactual regret minimization wasn't invented until 2007, despite being a relatively simple algorithm to describe and all the essential intellectual building blocks being known since the 1960s.

von Neumann had invented extensive form games with the express idea of modeling poker; there were researchers in the 1950s/60s (Hannan, Blackwell) working out the core ideas of regret minimization. One innovation (the sequence form representation of game strategies) was only known in the Soviet Union in the 1960s and not propagated to the West. But this was independently rediscovered in the early 1990s and it still took another 15 years for CFR to be developed.

Can you please give a link to the soviet paper? Thank you very much :)

The title is "Reduction of a game with complete memory to a matrix game" by I. V. Romanovsky, Doklady Akademii Nauk SSSR 1962. Unfortunately I don't have access to the article.

Thank you!

Reply on news.ycombinator.com