Artificial Intelligence in Games News – Week 11
A lot of this weeks artificial intelligence news comes from China
Welcome to the weekly roundup, where we draw your attention to everything newsworthy in artificial intelligence in games! If you’ve found or shared something that you think belongs here, contact us on Twitter! Last week, we covered Bixby, The April Fools roundup and a weird Pacman glitch? You can find that here.
DeepMinds Artificial Intelligence Heads To China
Later this month DeepMind, are taking Alpha GO to China in order take on more of the world’s best players. This news comes after the AI defeated Lee Se-dol back in March 2016, although many did not see him as the true worlds best. Instead all eyes will be on the showdown between computer and 19-year-old Mr Ke. It’s thought by many that Mr Ke is still expected to lose as Alpha GO has been playing online against professional players and is still yet to be defeated. This is largely due to the unusual yet effective moves it employs to surprise its human counterpart.
America’s Poker AI off to China
In other ‘AI destroying humans at our favourite games in China‘ news, we have Lengpudashi. Earlier in the year you may remember we wrote about Libratus, the AI responsible for convincingly beating some top poker players. Now the team from Carnegie Mellon University have powered the AI in robot Lengpudashi which translates into “Cold Poker Master”, an upgraded version of Libratus. In total around 36,000 hands will be played over 10 days, significantly less than the 120,000 played just months ago. For those not in the know, a successful poker playing AI is impressive because unlike GO there is hidden information, in other words players could be bluffing, making things much harder for the AI!
Starcraft One Step Closer To Biting The Dust
We’ve often discussed what game will be next to fall at the hands of AI and once again the finger seems to be firmly pointed at Starcraft. Oddly enough, the exciting news continues to come from China as a team from company giant Alibaba have published a paper outlining a system that can successfully organise combat by employing complex strategies learnt from the pros. It uses a trial and error approach removing the need for supervision which makes it even more astonishing.
80s Algorithm Beats DeepMind
OpenAI, one of Elon Musks many innovative projects has recently had success by revisiting neuroevolution. More specifically evolution strategies, or algorithms for solving optimization problems have been used to trump DeepMinds achievements in game playing. OpenAI developed an AI that managed to master Atari games in one hour compared to DeepMinds one day. The secret isn’t in how smart the AI is but rather how quickly it can try and fail using a significantly parallelized approach. Certainly an interesting read!