Explained simply: How DeepMind taught AI to play video games

Aman Agarwal, Medium, Aug 28, 2017
Commentary by Stephen Downes

I think the concept here is a good idea - take a complex paper on AI and rewrite it so most people can understand it. And it more-or-less works. We certainly get a sense of the accomplishment: they trained an AI to play different games by using only the game as input - no rules, no hidden data, just game input. But this post uses the phrase "that is self-explanatory" far more frequently than such a post shoud (a dozen times, at least). And most of the time the paragraph that the author thought was self-explanatory wasn't. So while I give the post high marks for idea, I would grade it much lower on execution. And yet - it's still worth a read, especially if you think mmachines can learn how to do complex things all by themselves.

Views: 1 today, 395 total (since January 1, 2017).[Direct Link]