Subscribe to the Sports Tech Feed

Complete your details below to get the latest news and content from the global sports & technology industry

Using ESports to build better A.I.



A.I. has already proven to be smarter than humans in games. Machines in the past have beaten humans in chess, Go, Ms. Pacman and poker. But what about ESports?

OpenAI has trialled and tested just that. On Saturday, the 13th of April in San Francisco, the non-profit held an event to showcase the potential of A.I. to succeed in ESports. A.I. battled skilled players from the best professional ESports team in Dota 2, one of the most popular online strategy games. In the best of three game match, A.I. prevailed as the champion. Winning Dota 2 is no small feat. The game is involves carefully planning attacks, defending towers from enemy attacks and distributing items to armies of heroes. The number of possible interactions between heroes, enemies, allies, and inventory can be overwhelming even for the smartest humans.

As Fortune reported, the goal of this event was to showcase the potential of deep learning, a type of A.I. used for recognise objects in photos, and reinforcement learning, a technique in which computers learn and build upon existing knowledge by trying different scenarios. The event was promoted by OpenAI, a group that not only organised the competition but also created the winning AI technology that won against the human players.

Open A.I. was founded back in 2015, with investments from big names like Elon Musk, the CEO of Tesla, and Reid Hoffman, the co-founder of LinkedIn. They dreamt of achieving artificial general intelligence (AGI), the point at which computers become as smart as humans. The use of video games to improve AI is not new. They have long been a tool for researchers to enhance A.I. for everyday tasks. For example, AI techniques used in video game simulations can be used in shopping apps to make better product recommendations to users. OpenAI co-founder and CTO Greg Brockman says that the Dota 2 win shows how AI can be used in tasks beyond its currently imagined possibilities.

With the growth of ESports, it looks to be a promising space from which A.I. can continue to learn and develop. Video games like Dota 2 are fast becoming a benchmark in A.I. research. However, the process of data training A.I. systems is time-consuming and expensive. Although OpenAI has declined to disclose how much their A.I. systems cost, they revealed that it took 10 months to train. OpenAI recently created OpenAI LP to make fundraising easier. Brockman say, “we will need to raise billions of dollars. This system is not free.”

Leave a Reply