OpenAI Five beat the World Champion at their own game


If you remember, in 2017, OpenAI, an AI startup backed by Elon Musk beat a world-famous Dota 2 player in a 1v1 match. With that OpenAI’s aim was to compete in a full-on 5v5 game. Well, that aim was met just a few days ago. OpenAI Five beat Team OG 2-0 in a best of 3.

Oh how the mighty have fallen

In case you didn’t know, Team OG took the No. 1 spot at The International last year. They walked away with prizes amounting to $25 Million. Aiming to battle with the best of them, OpenAI Five challenged Team OG to a best of three match to see who would win.

OpenAI Five
A fan art of Team OG representing the heroes they played (Image Credits: ArtStation)

The OpenAI Five bots were all trained using the same deep reinforcement learning methods and were controlled independently by different layers. The principle behind reinforment learning is essentially a trial-and-error approach. The AI would be dropped into an environment, having no prior knowledge about its surroundings. It would learn by means of a reward system and other incentives.

OpenAI Five
OpenAI Five uses reinforced learning (trial-and-error) Image Credits: Medium)

According to Greg Brockman – CTO of OpenAI, OpenAI Five improves itself by playing against itself in an accelerated virtual environment. Just to put things into perspective, within its 10 months of being developed, OpenAI Five has already racked up close to 45,000 years of Dota 2 gameplay.

The Battle Begins

Rather than pick from over 100 playable characters that Dota 2 offers, OpenAI limited the number of heroes used to 17. In keeping with tournament rules, both OpenAI Five and Team OG chose the Captain’s Draft” game mode. Here, each team is allowed to pick and ban one hero.

OpenAI Five
Captain’s Draft allows a team to pick and ban a hero one at a time (Image Credits: Reddit)

This is to level the playing field between both teams. It also means that the team has to be extremely intuitive in who they pick and ban. There would also be no summoning or illusions. So hero copies would not be available. This was because OpenAI hasn’t trained OpenAI Five for it.

Taking Dota 2 Bots to a whole new level

Match 1 of the tournament saw OpenAI Five delivering an aggressive push tactic. One eyebrow raising decision that OpenAI Five made was to buyback a character into the game. Buyback uses a chunk of in game currency to instantly revive a character (in this case Gyrocopter). Buybacks are usually meant as a last resort when the enemy team is close to victory.

OpenAI Five
Gyrocopter buying back into the game was unexpected so early in game, but it worked out for OpenAI Five (Image Credits: LiquidDota)

But in this case, the early buyback by OpenAI Five gave them an edge that OG could not overcome. The result: OpenAI Five won the first game.

Match 2 was perhaps more brutal. OpenAI Five seemed to have learned a lot from the previous match and continued its onslaught against Team OG. While Match 1 took around 30 minutes to complete, Match 2 was done in almost half that time.

Is OpenAI Five the future of gaming?

Now before you start panicking and throwing in the towel, this is not the end for humans to take part in competitive gaming. Rather, it’s the beginning of an interesting collaboration. The research team at OpenAI will no longer have any public demonstrations of OpenAI Five. Rather, they are working on software to allow humans to collaborate with OpenAI Five in real time.

What this essentially means is, rather than battling against 5 bots, you can actually play alongside OpenAI Five. The bots would learn the strategies you use and even come up with strategies of their own. To that extent, OpenAI has also released a platform to play against OpenAI Five.

OpenAI Five
Will Bots play our games for us? Or pose a more difficult challenge? (Image Credits: PointsPrizes)

According to Sam Altman – Co-Founder and CEO of OpenAI, platforms such as this could result in vast benefits in the future. He added that the platform would enable “AI working alongside humans to make humans better and have more and more impact”.

Altman went on to say that “what OpenAI is trying to do is build general artificial intelligence and to share those benefits with the world and make sure it’s safe”.

It’s pretty safe to assume that OpenAI will soon be able to play a wider variety of games. That’s also not such a bad thing. Just imagine the next time you load up a multiplayer game and your team is comprised of intelligent bots? Now that makes things a lot more interesting. You can learn more about OpenAI Five here

What do you think of the future of gaming? Would you play against or with a team of OpenAI bots? Leave a comment below.


Please enter your comment!
Please enter your name here