Queue all the Skynet jokes. An artificial intelligence (or AI) developed by British tech company DeepMind Technologies has gotten so good at StarCraft 2 that it’s already winning against pro players.

In an exhibition today, DeepMind showed what their AI “agents” can do when faced with some of the game’s most experienced veterans. And what it can do is beat them rather convincingly. The developers showed footage of Team Liquid players Dario “TLO” Wünsch and Grzegorz “MaNa” Komincz taken on the AI only to lose 0-10 in PvP mirrors.

But how did it do it? How did a piece of software defeat the experienced and adaptable mind of a real life human?

First, the AI (which DeepMind call AlphaStar) starts observing human StarCraft 2 replays and through imitation learns basic concepts like micro and macro. Then, different AI agents are thrown into a league of sorts, where they start playing each other. With the growth of the league in terms of participating AIs, the agents started developing strategies, counter-strategies, and counter-counter-strategies. By using Google’s TPUs (tensor processing units) and playing for two weeks straight, each agent “experienced up to 200 years of real-time StarCraft play”, the developers wrote Here, you might be thinking that the AI beat MaNa and TLO because it could move faster, but that’s not the case. According to DeepMind, AlphaStar had a lower mean APM compared to both players.

But even such advanced AIs have limitations. So far, AlphaStar has only learned to play Protoss and in a live exhibition, MaNa proved the AI can be beaten. But DeepMind’s goal in the end is to create something that goes beyond just beating players at a video game.

“The fundamental problem of making complex predictions over very long sequences of data appears in many real world challenges, such as weather prediction, climate modelling, language understanding and more,” the developers wrote in their blog, and added:

“We also think some of our training methods may prove useful in the study of safe and robust AI. One of the great challenges in AI is the number of ways in which systems could go wrong, and StarCraft pros have previously found it easy to beat AI systems by finding inventive ways to provoke these mistakes.”