Artificial Intelligence bots defeating humans in games is not new. There have been multiple cases where AI bots have defeated world class players at games like Chess and Go. However, the Elon Musk and Sam Altman startup OpenAI, which studies artificial intelligence has managed to beat multiple world-class DOTA 2 champions in one-on-one games played recently. The DOTA 2 video game was developed by Valve Corporation and the event was one of the biggest eSports events in the world with many players and fans tuning in to watch the AI vs Human matches being played. The results were not exactly surprising as most of the matches between humans and AI have seen the latter win, but considering that DOTA is a lot more complicated, the win has generated substantial interest. Also Read - Microsoft replaces human journalists with Artificial IntelligenceAlso Read - How AI will change the way social products are experienced
There were a total of five rounds the bot was scheduled to play against Danylo “Dendi” Ishutin, who is currently regarded as the best international DOTA 2 player in the world. The best of five would win. The bot managed to win the first rounds hands on. The second round was more complicated but Dendi resigned from the round. Reports suggest that the world champion refused to play the third match against the bot. Many other players too were defeated by the bot as a part of the championship. However, the matches were not played according to the standard DOTA format which requires there being two teams of five players each, with each team occupying and defending their own place on the map. Bot played the matches one-on-one with the players. Also Read - Elon Musk has an AI at Tesla that reports directly to him
According to OpenAI, the bot was trained using thousands of lifetime of matches. It played the matches against itself as part of the learning and practice process. Creators also claim that the bot was not taught any pre-programmed moves or any imitation learning techniques. All the learnings for the bot were done in matches and in sessions competing against itself and trying to beat the previous record or game. “This is a step towards building AI systems which accomplish well-defined goals in messy, complicated situations involving real humans, the company said. The total training period for the bot, however, did not exceed two weeks as opposed to Dendi’s years’ worth of training and winning matches. Dendi himself acknowledged that the bot played like a human but at the same time felt somewhat different as it was a machine. ALSO READ: Elon Musk s Tesla now onto self-driving semi-truck, aims for trials in Nevada
The bot might have won this match but the larger goal of the company is to continue training the bot till it manages to win against the tradition 5-member team players in challenges. The final goal is to comprise a mix of human and AI bots in each team and pair them off each other. Despite the achievements of the bot, Elon Musk seems less than cheerful as he compares the artificial intelligence bot to the fear from North Korea. In a subsequent tweet, he voiced his concerns over safety and privacy considering bots are soon becoming the norm of the day with most technology companies experimenting with artificial intelligence to some level. Musk has been previously vocal in voicing his concerns regarding AI and pushing for a regulatory body set u to monitor AI’s progress.