Yesterday at a showmatch in San Francisco, champions of The International 2018, OG, lost against OpenAI Five.
As mentioned in the previous article, OG, one of the world’s best Dota 2 team faced OpenAI Five, a team of five AIs, for a competitive 5-on-5 Dota 2 match. The showmatch, which was later called the OpenAI Finals, took place in San Francisco at 11 PM IST yesterday.
The OpenAI Finals
It was a best-of-three contest, with OpenAI Five winning two matches back-to-back. The match was, however, played with certain limitations, which would eventually turn into a big disadvantage for OG. Each team was only allowed to choose from 17 heroes instead of all the 115 characters in the game. OpenAI Five also made use of ‘Captain’s Draft’, a game mode which allows each team to ban heroes which would prevent their opponents from choosing those characters. Summons and Illusions were also disabled. For those people who do not know what these are, Summons allow players to create and control additional units, whereas and Illusions allow them to create copies of themselves. With these features disabled, OG basically had to play the match with one hand tied behind their back.
OpenAI Five relied on the strategy of aggressive advancing. In the first match, they pushed back OG using aggressive maneuvers and upon the death of one of their heroes, they instantly revived him even though their base wasn’t under attack. The match continued for a little more than 30 minutes. During the second match, OpenAI Five overwhelmed OG as they ruthlessly advanced towards OG’s ‘Ancient’ without giving them a chance to strategize their defense tactics. They won the match in about a little more than half the time they needed for the first match.
You can watch both the matches below –
Strategies used by OpenAI Five
As a team consisting of Artificial Intelligences, OpenAI Five had superior reflexes when it comes to ‘micros’, a word used to refer to second-to-second positioning and attacks. They knew that the longer they drag the match, the lesser their chances would be for winning. Therefore, they advanced aggressively and finished off the match as quickly as possible without giving OG a chance to reinforce their defenses. An AI is always good at deciding the best possible approach to any given situation which is why OpenAI Five used buyback strategy upon the death of one of their heroes.
However, it is to be noted that OpenAI Five is given access to the same amount of information as humans. But since AIs have much faster response time than humans, they are quick to point out data such as positions, healths, and item inventories that humans have to check manually.
“OpenAI Five is powered by deep reinforce learning, which means we didn’t code it how to play. We coded it how to learn,” said Greg Brockman, Co-Founder and Chairman of OpenAI. “In its 10 months of existence, it’s already played 45000 years of Dota 2 gameplay. That’s a lot — it hasn’t grown bored yet.”
Before the contest, the commentators stated that the game looked evenly matched. OpenAI, however, declared that they calculated a 90% success rate for their team. OG was really excited as the showdown would prove the superiority of humans or AIs. However, their defeat against OpenAI Five changed that. After the second match, team captain Johan ‘Notail’ Sundstein said, “It hurts. We’re doomed.” He later tweeted, “Just hope they remember how nice and mannered we were once they own the planet.”
Towards the end of the event, the OpenAI research team announced the release of their platform named the ‘OpenAI Arena’, in which players would be able to play against OpenAI Five as well as play alongside AI bots and learn new strategies from them.
In an interview with The Verge, OpenAI CEO and Co-Founder Sam Altman said, “That’s an important lesson for how the world is going to work, training these things and have them work in parallel. Collaboration is one of the more positive visions we have for the future of the world — AI works alongside humans to make humans better and have more fun and more impact.”
While OG fans are disappointed at their defeat and worried that AIs might takeover the roles of human beings, the OpenAI research team has stated that contrary to popular belief, AIs can actually be used to benefit us. Their accuracy and precision at solving problems will help us with our decision-making capabilities and give us new solutions to peculiar problems. OpenAI CTO Greg Brockman stated that Artificial Intelligence can transform the society in the same way as internet did.
I guess we can all agree that, if used properly, AIs can revolutionize our approach to modern day problems and can indeed be used to train us to be better and more capable.