Image by OpenAI
“We predict the probability of winning to be 95%” echoes the OpenAI bots in the opening minutes of the match. This weekend in an exhibition match, the OpenAI Five team dominated five pro level players in an exhibition match.
Blitz, former caster and pro player, said that the match felt like no opponent he had played for. It felt like playing a strong team, where each small error compounded over and over. This opinion was a testament to how far the OpenAI project has come since it’s one on one tournament at TI last year.
The comparison of OpenAI to a really good opponent was enlightening. More than being an oppressive opponent with mechanical advantages, removed from limits like the rate of human action and thinking, OpenAI now feels more and more like a thinking being. They react in realtime. They seamlessly intermesh their spells, weave between farming and pushing strategies, and group at the most opportune times to push.
Playing against five audience members, the OpenAI Five decimated their opponents outer towers and raxes in about thirteen minutes. In their opening 2-0 against Team Human, the games averaged 23.5 minutes, but the outcome was determined long before.
Even Team Human’s small victories—a first blood here, an early Shadowblade on Shadow Fiend—felt like it was little recourse. It felt fatalistic, as if OpenAI let them have those crumbs, because in the end it didn’t matter. The win probability would nonetheless rise, rise from 90%, to 95%, to 99%.
To be fair, the game wasn’t without its restrictions. The pool was limited to 18 heroes, expanded from 5 in the initial announcement. Each team had 5, indestructible couriers—just a layover from the OpenAI 1v1 code. And the OpenAI bots had an artificial 200ms delay, which was partly an artifact to allow the team to scale their project, but it also happens to be close to a human’s reaction time, even if it seemed as though Fogged’s Earthshaker was instantly hexed on his blink initiation.
The bots are learning from playing 180 years of matches, per day. They recalibrate on the spot, dissecting which actions lead to the most advantageous path to victory. One audience member asked if the OpenAI bots would learn how to drop certain items while consuming regeneration items to maximize health or mana recovery--one of the numerous tricks that high level players do in order to extract every advantage they could. The team simply deferred, saying that yes, if the OpenAI bot felt like it increased their chances to win.
What players might learn from watching the bots, rather, is how good a team can be when it’s deliberate in its decision making. To borrow from basketball’s parlance, it’s to play with “thrust.” Make quick decisions, and then execute. It’s possible to push towers without pushing heroes, if you just get your team to attack the tower together. It’s possible to dominate a lane with three supports, as OpenAI did when they trilaned with Lion, Crystal Maiden, and Lich. Perhaps one of OpenAI’s greatest strengths is they haven’t yet learned how to flame each other.
The OpenAI bots didn’t even last hit that well, but they were near perfect in their spellcasting. They never overlapped stuns and they always found a way to nail AoE spells at their maximum possible ranges. They maximized their mana pool every fight, even spamming Assassinates and nukes before a team fight broke out. Capitalist’s Crystal Maiden couldn’t even stand in the lane, even far behind his own tower.
And then there’s the importance of the draft. In game 1, OpenAI calculated a 95% win probability solely from the draft, 76.2% for game 2, and in game 3, the OpenAI devs, in a spell of overconfidence, allowed Twitch chat to draft their final lineup. The crowd cobbled together a team that the OpenAI Five predicted would give them a 2.9% chance of winning.
The next step for the OpenAI team is removing some of their imposed restrictions, smoothing out some quirks like warding and item buying, and finally playing a pro team at the upcoming TI. The question may not be how relevant Dota can be if computers can best humans at it, in real time. Chess and Go have been able to thrive despite Kasparaov and Lee Sedol falling victim to Deep Blue and Deepmind, respectively. Already, we’re seeing glimpses of the project’s real world applications. The foundation of OpenAI Dota helped the team program a robotic hand to mimic the dexterity of a human hand. Or to be more accurate, the team trained OpenAI to do so. What more can this machine learn?
bots monkaS
THIRST! 3st
Report AI for communication abuse
Brace yourself, Skynet is coming
It's nice, but their hero pool still really small and they can't play against illusions, I mean it's not even Dota...
I'll be impressed when the restrictions are removed.
imagine if openai can play meepo.
monkaGIGA
who cares lol
People are not impressed with the limitations.. chill,it's just in beta stage 2 mode.. maybe in a few years time it could go even further who knows what the sky is their limits.. give it time to learn.. but I do say I'm not impressed with how draft depended their are,what I would like to see if they come from behind..
Can't wait until OpenAI mastering IO and Earth Spirit.
@stars The devs said playing with illusions would be an unfair advantage for the bots due to their superior micro skills. Therefore it's not that they can't play against illusions, but just that the devs have prioritized other aspects of the game over that one so far.
One day, you filthy humans will not be able to beat OpenAI.
“With the current toxic Dota 2 community, we predict the probability of winning any games to be 99.99%”
Humans would wipe them with all heroes and illu, the bots have practiced those heroes 1000 times more than any human but only against those other 14 heroes.
Ai bots vs team team pro players 1 miracle 2 noOne 3general 4gh 5jerax
There’s still infinitely many things before the bots will be able to defeat a team without any restrictions. They still have to learn the rest of the hero pool, all the potential interactions and movements... merely the concept of the draft, with countless factors to think about. throw in heroes whose design are insanely difficult to comprehend at the level they want these bots to, such as any of the micro heroes, illusion heroes. Throw in the concepts of stacking and pulling. I give at least 2 years before Dota ends, and Dota will end when they create a bot team capable of guaranteeing victory against a human team
who the hell is those team human ?
It's an interesting concept. And i doubt dota will be dead when they finaly pull it off. they have no intention on selling the bots or whatever (maybe to valve for optimizing their own bots) but this is a company that doesnt give a shit about esports. they're only interested in the behaviour of their AI in a complex enviorment
KI will Dominate the World soon, and we must think about Laws and restrictions about it or we will end in a Future painted in nomerous Sci Fi Movies.
So every teenage fuck in the comments decided to talk about how unimpressed he is about the bots, and how much better is he and his archon buddies at dealing with illusions. Do any of you have the slightest clue about Machine Learning and AI field? OpenAI bots are truly groundbreaking, and are only possible recently due to insane processing powers available. Do you think that devs and sponsors care about your kiddy game? This is merely a step in learning to build complex networks, a model with incomplete information in a controlled environment. This is a fucking set of 1’s and 0’s that through trial and error taught itself to play dota better than any of you brilliant minds. This shit is going to change lives and whole industries, yet you are not impressed cuz it abuses couriers
why do i have that weird greek pantheon like thingy next to my name
AI is just better than you
It's fucking obvious humans can't even compare to AI.
Humans can't make perfect decisions, and they have latency enough for them to always be the ones that get hit by spells and not vice versa.
I don't even know why anyone would expect anything else. No matter who Team Human consists of, they will lose, as long as the AI is taught what they are playing against.
You AIs may be better at dota than humans, but who made you. YES! whose smarter now.
wasnt there a 3rd game, though? slightly over 30 min long, won by humans.
besides that, human team wasnt, hmm, a stack of tier1 players. afair it mostly consisted of casters (merlini, capitalist, etc.)
Amazing, you mean to tell me a computer is better than a human at playing a computer game???
Well they defeated a team full of top 500 players so its pretty impressive. also i would like to see their hero pool expand to over 50 heroes, its unrealistic for them to have over a hundred heroes learnt in 2 years.
I will be more interested in the outcome of these games when they start using 1 courier which can be killed on each team like in real Dota.
Allowing each bot to have its own invincible courier kind of defeats the purpose of simulating teamwork in my opinion. The pace of the game is completely different, there is no tempo or strategy all they do is spam regen constantly and overpower their opponents with basically hacks (perfect reaction time and mechanics).
Am I surprised bots that process every single variable on the map instantly 60 or more times per second can win mechanically in team fights? Not at all, they really need to fix this courier issue so that the bots actually need to MAKE DECISIONS AS A TEAM not just having every bot individually "hack" their way to victory in flawless team fights and no down time.
it's not about the win we are going to be impressed, but to make AI plays DotA even merely herald level normally (naturally) is already a great success!
I believe it's not something ever close to easy to be pulled off.
But these guys trying to point out any factor in the game (Courier, instant reaction mechanics etc) do have a point, let's say we can access bot game with such AI in a normal DotA would be very satisfying (current bot buys wrong item components)
"Perhaps one of OpenAI's greatest strengths is they haven't yet learned how to flame each other".
You know what to do boys, teach those bots! :D
why the hell is the video autoplaying, even when not scrolled?
Gonna wait till AI learns how to combo with Invoker so we can see that juicy perfect combo + refresher orb :))
Don't let OpenAI invest an AI on Meepo, ever. I believe it would be like facing 8 players or 9 with Aghanims.
There is one thing bot will never have and that is imagination i would love if we could also play vs bots, that human team looked like they have never played together before