AI dominates professional Dota 2 match

Dota 2 match AI

The strength of an advanced system, OpenAI Five, was demonstrated at a live event in San Francisco, where OpenAI Five competed against a team of human professionals in Dota 2.

The exhibition emphasised the extent to which artificial intelligence has advanced to excel in the complex and demanding real-time strategy environments.

OpenAI states that the system is very strong because it is able to play the game continuously, in effect where it can play the equivalent of 180 years of gameplay in just a single day.

This intensive training enables the AI to hone its tactics, coordination and decision making at a rate that no human team can match.

The human lineup included famous and seasoned players, such as Blitz, Cap, Merlini, MoonMeander, and Fogged.

Although their skills were on the level of a professional, they were not able to match the accuracy of the AI.

During the first game, OpenAI Five was able to gain control of the game quite fast, thanks to its perfect coordination and map knowledge, which forced the human team to concede only 21 minutes into the game.

The second game was a bit more competitive. Humans started finding minor flaws in the AI gameplay, and managed to defend themselves in the very beginning.

Nonetheless, towards the end of the game, the AI improved, recovered its pace after the 20-minute mark, and won another unquestionable victory at 24 minutes.

The third game had a very interesting twist, wherein spectators on Twitch had the privilege to select the heroes of the AI.

Having a poorly optimised lineup, the AI was finally beaten, and the human team was easily defeated.

Players subsequently reported that the main strengths of the AI were its immediate decision-making and extremely aggressive strategies.

Although weaknesses such as slower reactions to invisibility and imperfect placement of wards were detected, they were not sufficient to always defeat the system.

Also read: New study finds AI can now detect pancreatic cancer years before diagnosis