Main menu
You are here
Go (board game) AI
This Thursday, April 14th at 7PM in Dreese 264, Daniel Thau will be giving a presentation on open source artificial intelligence for the board game Go.
Go is an ancient board game (estimated to be 2000-4000 years old) which, despite a simple rule set, is extremely difficult to write a human-level AI for. Unlike comparable board games such as Chess, the best Go AI's in the world were not match against strong club players only a few years ago, and despite significant development recently they are still no match for weaker professional Go players today.
Daniel will be discussing why humans are so much better than computers at Go despite the simple rule set and two general approaches taken by open source Go AI developers.
The talk is intended to be interesting to seasoned Go players and AI hobbyists as well as those who know nothing about about the game or AI development.
| Attachment | Size |
|---|---|
 goaitalk.pdf | 432.41 KB |
 goaitalk.tex | 40.27 KB |
Comments
Slides
The slides, both as PDF and the LaTeX source code, are attached. It was a bit rushed but the LaTeX is fairly clean, so if you want to browse the code it shouldn't be to bad of an experience. The TikZ framework for rendering the go boards is kind of snazzy - it has a few tricks I didn't employ for the presentation due to time constraints.
I fixed one typo (Chose -> Choose), but baring that they're exactly identical to what I used in the presentation. I manually checked all the hyperlinks in several PDF readers from several different Linux distributions and they all worked fine, so I'm not sure why some of them failed during the presentation. If they don't work for you, please let me know what PDF reader, Internet Browser, and Operating System you are using so I can root out whatever the cause is and ensure it doesn't happen in future presentations.