Back in June, OpenAI was discussing their progress on creating AI bots that could effectively beat a team of pro Dota 2 players at this year’s “The International 2018.” With the event having taken place over the weekend, the answer to who would triumph between man vs. machine was answered… looks like we came out on top this time!
OpenAI, a company funded in part by Elon Musk, first tested their AI against Dota 2 players at last year’s “The International 2017.” At the time, the AI bot was tested in a 1 vs. 1 match against Dendi, a pro Ukrainian Dota 2 player. That day was a sad one for mankind as the bot came out on top
At this year’s The International 2018 which occurred over the weekend, the battle of man vs. machine continued in the same best out of three format, except this time it was a full 5 on 5 match. Given the increased complexity teamwork brings, were the bots smart enough to adapt to different team strategies to claim victory? It seems that the day of man isn’t quite over yet, as a Brazilian team and one Chinese team beat the bots in back to back matches.
According to PCGamer, the bots had a hard time dealing with mid-game roams by human players. While the bots could maneuver really well in team fights leading to an almost equal number of kills, the gold difference still managed to sway in favor of the pro players as they were able to do a better job farming up. Similarly, pro players picked up on certain ways of forcing the bots into abnormal situations which they had no idea how to deal with. One particular strategy that PCGamer pointed out in their article is this hilarious one.
Dealing with cheese/trolling techniques is something that seems like bots would have a hard time with. Despite this win, it’s probably still only a matter of time before the AI bots get the better of us. Given that the series ended right after the back-to-back wins by the pro teams, we don’t really know who would’ve won if a third match had been played. Hopefully, it still would’ve been the pros!
Featured Image Source: Dota Dose (YouTube)