I'm working on a project that's been going on at least 10 years, not sure exactly how long. It's an engine to play the game of go that uses something like genetic algorithms to produce generations of "go-bots". I have sets of go-bots fighting each other for dominancy pretty much 24/7 now. Deepmind and alphazero have nothing to worry about, though. I play them against the gnugo program as a test of their worthiness so I can see if they are improving or not. The best of them can win 1 or 2 games in a hundred against the gnugo program even when given 20 stones to start with. They don't actually look ahead any moves, though, they just look at the state of the board and figure out where they want to play next based on weightings for different things, so it's maybe not as terrible as it sounds. I'm pretty sure that the top go masters are sleeping like babies not worrying about my program at all, though.<p>Anyway, I didn't start this as a project I was going to sell or even one that had a defined deadline, it's a hobby. I've learned lots of things along the way. I think of it as a nice wooden contraption I'm building that I sand and polish as I go. I recently (early this year, I think?) rebuilt my bots dna based on what I'd learned about how the last batch fared. Every day I come home and see how they did against gnugo and make sure the scripts have enough commands to perform so I can let them do their thing without intervention. I work on other projects but still keep this one going because I enjoy it. My "goal", such as it is, is to evolve a bot that can beat gnugo more than 50% of the time on an even playing field, without looking down any sort of move tree.<p>So this is my shed, I guess.<p>edit: grammar