DOTA 2: Professional players beats AI Machine

So far we have arrived, machine. You have defeated chess or Go , but (at the moment) you have not been able to give us DOTA 2 .

That is what several professional DOTA 2 players could have commented on after defeating the OpenAI team that had developed an artificial intelligence system to try to win human players to this strategy game. In a tournament with former Chinese professional players to the best of three games the humans won 2-0 to the machines. Hurra .

The long-term strategy fails

The team of five human players played against that team of machines driven by an artificial intelligence project called OpenAI Five . They did it in the well-known tournament The International 8 , in which there is a total of 11 million dollars for the absolute winner.

In DOTA 2 the strategy is different than in games like chess or Go, and there is an intuitive component even more difficult to drive in those artificial intelligence systems, which despite everything put human players in trouble.

In fact the OpenAI Five team managed to get ahead on the scoreboard in the first few minutes, but the team of human players managed to recover and end up winning both that game and the next.

The rules for these games were somewhat more restricted as a somewhat more limited number of heroes and abilities to choose from, something machines compensated for with their instantaneous reaction times . Although the engineers tried to adjust the system so that their reinforcement learning system - which allows them to learn from mistakes and seek rewards for their actions - the second game ended in another defeat.

The programmers of OpenAI Five detected that these short-term rewards meant that the final objective (to win the game) lost some weight, which prevented the machines from having a longer-term strategy . That has meant that now these developers want to focus on that approach to the problem, and we will see if in future confrontations these machines effectively defeat human players.

For the moment, yes, we are still breathing something more calm . The machines have not exceeded us in this. Yet.