On Sun, Mar 13, 2016 at 12:33:55AM +0000, Gonçalo Mendes Ferreira wrote: > The single machine version and the 2000 machine version is apparently a > difference in 150 ELO, so maybe a 32 core instance in Amazon would be > close enough, costing about 1200$ a month. Maybe it's doable, or for a > Tokyo U. or BGA or AGA to set up something like this with Google permission.
It would be great if Google was willing to do this, but I'm not sure this is realistic at all, as it might be all built around internal Google frameworks and keeping it running at Google would mean someone has to take care of it. So I suspect the Go world has to be a little more patient until Deep Zen, Aya, oakfoam, possibly https://github.com/Rochester-NRT/AlphaGo and others catch up. Save the money for the openly available reinforcement learning iterations! Just to clarify, 4xGPU, 32xCPU instance is $1800/month, while AlphaGo single machine version uses 8xGPU, 48xCPU (and not sure what GPU generation). So that's another computation power halving. -- Petr Baudis If you have good ideas, good data and fast computers, you can do almost anything. -- Geoffrey Hinton _______________________________________________ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
