OpenAI at The International 2019: Bots vs Champions

OpenAI, the company that develops cutting edge artificial intelligence tools, got involved into Dota 2 around two years ago. Initially, they developed a bot that was able to beat any professional player in a 1v1 (Shadow Fiend vs. Shadow Fiend) duel. Then they built an entire bot team that got good results initially, but ultimately failed to defeat the TI 8 participants.


© OpenAi

OpenAI’s recent results

Around a week ago, OpenAI’s bot resurfaced with new and improved capabilities. And thanks to everything it has learned over the past 8 months, it crushed the TI 8 champions, OG, without even trying.

This year, we’ve also had another artificial intelligence system demolish one of the best StarCraft 2 players in the world. This player was TLO or The Little One and he lost 10 games in a row against an AI bot called AlphaStar. The name may sound familiar to you if you have an interest in chess, as the best AI bot that’s currently available in that game is called AlphaZero.

So overall, artificial intelligence is getting better and better at mastering closed systems and some of the areas where it’s been prolific has been esports.

TI 9 predictions

The current state of OpenAI’s technology seems quite spectacular and the company wants to test it against the world’s best teams at The International 2019. Based on what I’ve seen during the OpenAI vs. OG match, the chances that the AI bot defeats even the likes of and Team Secret are at least 50%. In other words, I do believe that this technology will soon reach a point where it becomes unbeatable, just like in other games.

Implications for the future

The implications of AI technology are huge for the whole esports industry. Pretty soon, people might be able to start training against an AI bot that’s much stronger than any professional team. In turn, this will allow pro teams to become much better and master certain play styles and hero compositions simply by learning from how the AI plays and reacts to them.

In chess for instance, people have been training with assistance from AI for many years. And it’s extremely common for pro players to assess the quality of a move when trying to figure out the best way to approach a position by checking what the computer says about that move. In the not so distant future, I’m convinced that professional Dota 2 teams will do exactly the same and start searching for better lineups and ways of playing the game by having everything tested against the best gaming brain in the world: AI.