Stunning indeed because we always thought that machines would hardly beat expert humans at games but it happened over the weekend. Elon Musk’s OpenAI bot defeated an expert at Dota 2. Dota 2 is an online multi-player game where players use characters called heroes to fight each other. A winner is declared when one team takes down an “Ancient” (a building or any other structure for that matter) in the base of the opposing team. The OpenAI bot had just two weeks to practise and still came out on top of players who have had years of experience in what some see as scary. Even the OpenAI founder Elon Musk thinks systems like this pose a greater threat then even a nuclear armed North Korea.
[xyz-ihs snippet=”elon-musk-openai-dota2″]
In a YouTube video, OpenAI co-founder and CTO Greg Brockman said “Our bot is trained entirely through self-play. It starts out completely random with no knowledge of the world, and simply plays against a copy of itself — which means it always has an evenly matched opponent”. This means like other artificial intelligence systems, the machine takes up a task and continues to self-improve as it gathers more data on the task until it attains perfection. With respect to Dota 2, it means the bot kept playing and learning with time till it became as good as the best human players who have had years of playing experience and even went ahead to beat them in a remarkable feat.
[xyz-ihs snippet=”openai”]
Now this is significant not only because the bots are coming for us, it’s how quickly they can learn to do what has taken us years to accomplish.
The next step for the bot is international Dota 2 world championship. It kicked off in Seattle and the grand prize is $10.7m.
With the bot now at par with professionals like Danil Ishutin (aka Dendi who is a Ukrainian Dota 2 professional).
OpenAI won’t get the money though because it was more of a side attraction at the tournament but this could be the beginning of things to come in future.
Due to the processing power needed to carry this out, Elon Musk said they used Microsoft’s Azure cloud platform and publicly thanked them on Twitter saying “Would like to express our appreciation to Microsoft for use of their Azure cloud computing platform,” he wrote. “This required massive processing power.”
While they claim that OpenAI which is a non-profit wants to apply artificial intelligence safely, it’s not clear what safely means at this point. Google’s DeepMind is developing similar initiatives around gaming but has now decided to take it to real life solutions.
Discover more from TechBooky
Subscribe to get the latest posts sent to your email.