A competition pitting artificial intelligence (AI) against human players in the classic video game Doom has demonstrated just how advanced AI learning techniques have become – but it's also caused considerable controversy. While several teams submitted AI agents for the deathmatch, two students in the US have caught most of the flak, after they published a paper online detailing how their AI bot learned to kill human players in deathmatch scenarios. The computer science students, Devendra Chaplot and Guillaume Lample, from Carnegie Mellon University, used deep learning techniques to train their AI bot – nicknamed Arnold – to navigate the 3D environment of the first-person shooter Doom. By effectively playing the game over and over again, Arnold became an expert in fragging its Doom opponents – whether they were other artificial combatants, or avatars representing human players. While researchers have previously used deep learning to train AIs to master 2D video games and board games, the research shows that the techniques now also extend to 3D virtual environments.

Computer AI has long been defeating humans in the game of chess, and earlier this year Google's AlphaGo became the first to beat a world champion in the ancient Chinese board game Go, but now it's time to settle the score in a game that really matters: a classic FPS. That's what a group of AI researchers hope to see, with a new challenge inviting computers to face-off in Doom, but playing as humans would. The "Visual AI Doom Competition" will be hosted by the 2016 Computational Intelligence and Games (CIG) Conference later this year, and now the group is accepting applications to find the best bots that can play a round of deathmatch against each other using the same learning techniques as a human. But this isn't the same kind of AI used for enemies when you play a game against bots. No, this kind of computer AI will not have all-encompassing knowledge of the game and its workings, instead it will have to rely only on what it "sees" as input.