It looks like the days of humans having any chance against AI in games could be numbered, as Google's AlphaZero system of self-teaching AI game playing is becoming all too powerful. After the bespoke AlphaGo system beat the world Go champion after playing the game a few million times earlier this, the more general-purpose AlphaZero has now learned chess in just a few hours and subsequently beaten the world champion Stockfish AI in a best of 100 matchup.
In total AlphaZero spent four hours learning how to play chess before taking the fight to Stockfish. Over the 100 subsequent games, 72 of them were draws, but AlphaZero was able to win 28 of them, making it the victor overall.
There have been some concerns raised by Chess aficionados, most notably that Stockfish wasn't given quite the same processing power as it has had in the past and for whatever reason, did not have access to its typical list of opening moves, but regardless, AlphaZero is clearly the more powerful AI and will only improve in the future.
What's especially exciting about AI matchups like this, is that they aren't holding back. As Demis Hassabis explains in his writeup, when AIs face off against one another they don't have to hide their true potential as they might against a human player.
If you're worried such AIs will one day lead to the downfall of mankind, or at least beat you in your next game of Fortnite, don't worry too much - yet. Games like Chess are very simple in that the world is constrained and with a limited number of potential moves. Once you expand those AIs tend to find things much harder - but not impossible, so watch out.
