GOOGLE's DEEPMIND AI is done with crushing humans at Chinese board game Go, and has been given a fiddle to now play other games.
In a paper snappily-titled 'Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm' and published on arXiv, the DeepMind team revealed it has reworked the AlphaGo Zero AI to play Chess and the Japanese semi-equivalent Shogi.
Apparently, the tweaked AI can batter other digital opponents at both games thanks to being able to develop elite game playing skills within 24 hours, with DeepMind's boffins noting it "convincingly defeat a world-champion program".
This ability to develop these superhuman board game skills come courtesy of AlphaGo Zero's capability for teaching itself through reinforcement learning, a technique whereby machine learning algorithms can learn without the need to fist absorb a load of data provided by humans.
This is in contrast to the AlphaGo AI which required human input to make it a world-champion crusher.
So armed with just the data on the rules of Chess and Shogi, AlphaGo Zero can teach itself the strategies to beat skilled board game players. Given the rules of Chess and Shogi are similar and a lot less complex than Go, the AI soon developed top-notch board game skills.
Of course, credit does need to be laid at the doorstep of DeepMind's smart alecs, who have found a way to generalise AlphaGo Zero's software to work across different board games. This also hints that the AI could master other games.
DeepMind wants the AI to master online multiplayer games. But as these are not rules based per-say and rely upon all manner of complex things working in tandem, such as strategy, the ability to adapt on the fly, and memorising maps, humans still have the edge over robots.
But it's arguably only a matter of time until DeepMind figures out a way to overcome these issues, as the march of AI development is showing no sign of slowing down.
This is great for people who welcome such smart systems, but for robot-fearing tech messiahs like Elon Musk, such AI creation is likely to induce more paranoia over killer robots and job stealing droids. µ
C3-PO, R2-D2, BB-8 and other Androids
Helpful cyber vigilante gets short changed by customer services
...you know, now it's less confusing...
Firm will no longer provide updates for its first Android mobe