The AI research firm decided it was time to call time on OpenAI Five after the clever bot managed to beat the world champions at the wildly popular eSport Dota 2.

OpenAI Five managed to win two back-to-back games against team OG at an event that was streamed live from San Francisco. While other OpenAI systems have beaten pro-gamers before, this is the first time one has done so in a livestream situation.

The AI system gained the edge over its fleshy meatbag competitors thanks to having experienced some 45,000 years worth of simulated Dota 2 playtime, something even the best eSports wouldn't be able to manage.

But to do that a fair bit of compute power was needed; 800 petaflops a day over a ten-month period. That gave OpenAI Five a 99.9 per cent win rate over a previous bot that battled Dota nerds at The International Dota 2 event.

"The current version of OpenAI Five has been training continuously since June 2018, despite changes to the model size and the game rules (including some fairly large game patch updates and newly implemented features)," the boffins at OpenAI said.

"In each case, we were able to transfer the model over and continue training—something that is an open challenge for RL [reinforcement learning] in other domains. To the best of our knowledge, this is the first time an RL agent has been trained using such a long-lived training run."

But trouncing humans was only one part of OpenAI Five's capabilities, as the bot can also play nice with its human counterparts. It could play cooperatively and in an effective manner with real-life Dota 2 players, which OpenAI championed as "a compelling vision for the future of human-AI interaction".

Hopefully, that'll mean just cooperation and not any form of enslavement later on down the line.

OpenAI will be retired and an in-depth technical analysis of the tech will be released into the wild. But before that, there'll be a chance to pit your Dota 2 gaming wits against the bot through OpenAI Arena, which runs between 18 and 21 April.

And OpenAI doesn't seem to be done with developing AIs to tackle gaining: "We'll continue working with the Dota 2 environment within OpenAI," it quipped.

"We've seen rapid progress in the past two years on RL capabilities, and we think that Dota 2 will continue to help us push forward what's possible—whether with achieving competent performance from less data or true human-AI cooperation."

（By Roland Moore-Colyer From Theinquirer）