Elon Musk's startup OpenAI has defeated the world's best Dota 2 player using an artificially intelligent bot. For those that understand the complexities of #Dota 2 strategy, this is an incredible achievement for any player, let alone Skynet's Great Grandfather. The human vs. machine showdown took place at the annual Valve Dota 2 championships in Seattle USA, with Elon Musk announcing the achievement via Twitter.
Professional Dota 2 players practice for years to accumulate the necessary skill and experience needed as a contender on the pro tournament leagues. In a show of just how progressive OpenAI's bot truly it is, think about this: the botinator (I'm trying a few names for it) started playing Dota only two weeks ago...
This Is How It Works
Incredibly, the Open AI Benderbot (on the right track) trained itself entirely through a system of self-play. As explained in the above video the Halbot (still not the name):
Starts of completely random with no knowledge of the world, and simply plays against a copy of its self, which means it always has an evenly matched opponent, and it climbs this ladder until it is able to reach the skill level performance of the best professional players in the world.
As previously mentioned, this process of learning the 113 playable heroes in Dota 2 only started two weeks ago. Imagine what the I-bot (nah, that's not it either) could do if given room to truly explore its potential.
Machines One, Humans Nil
I feel sorry for Danylo “Dendi” Ishutin having to be the first player to tragically lose to a machine. In the one-on-one match he publicly stepped up and was defeated. To his credit, he gave it his all, so in that respect at least the human race was well represented.
It shows that even with skill, speed and intuition on his side the complex strategic implications of understanding all the cause and effect decisions proved the overriding difference.
The Next Step, A Full Team Match Up
The initial battle was an only the first introductory step for bot5 (too '80s?) to become alive. The next challenge for OpenAI is to push their Dota 2 project by building a team of five AIs to interact as part of a full team. If that is successful, they hope to then incorporate R2Dbot (that's the name!) with human players as part of a professional team. WTF!
We want to mix the AI's and human players on a single team and see if they can reach a levels of performance that neither of them could reach on their own.
What Will The Future Look Like?
In gaming terms, what happens if all this becomes a reality? The landscape of future esports could be a new frontier beyond what we once thought was possible. Will it be considered performance enhancing to include AIs similar to the Olympics? Or, will the competitive gameplay be at such a high level that audiences will see R2Dbots as a necessary step in the world of professional gaming entertainment?
Moving away from gaming, the even bigger question is what is the endgame for R2Dbot? Obviously OpenAI are using the Dota 2 championships to showcase their product's potential but it's a real eye opener on the strategic thinking capabilities between AI and humans. Hopefully the powers that be will think this through and direct its future research towards truly beneficial causes. I'm not keen on becoming a battery for any machine overlords any time soon.
Are you surprised about R2dbot's win?