OPENAI'S DOTA 2 PLAYING smart bots have lost their first rumble against professional human players at The International tournament.
In a best of three bouts between human pro players and OpenAI's team of bots, which are governed by an artificial intelligence (AI) system geared up for playing Dota 2, the first win went to the humans.
Two other rounds are scheduled and will take place against other human teams, so the bots could either dominate their puny human rivals or get given the run around by people with neurons and synapses governing their grey matter. rather than machine learning algorithms.
The bots came a cropper against humans from Team Pain, made up of five pros from Brazil, in a game that lasted some 45 minutes. The AI bots had a sketchy start but once they got into the game they proved they can go toe-to-toe with humans, evenly matching them at the beginning.
But in the latter stages, the bots looked to be dominating, putting their faster reactions into action and pulling ahead of Team Pain. However, that was short-lived and the humans made a comeback, pushing the bots back until they reached the AIs' 'Ancient' - the core structure of each team in Dota 2.
In the final battle of the game, Team Pain wiped out the bot player characters and then wrecked the Ancient, thus securing victory for humanity.
So while OpenAI's Dota 2 AI system and bots can beat armature players with suitable aplomb and show semi-pro gamers who's boss, it can't quite take on the true professionals, even in games that restrict the labyrinthine complexity of full-fat Dota 2.
Turns out the OpenAI bots seem to lack the ability to make true strategic decisions the way humans can, as mentioned by AI researcher Mike Cook on Twitter.
"The bots are still very good at moment-to-moment, but they seem bad on macro-level decisions," he said commenting on the game.
But I just want to reiterate: I cannot believe I just watched a team of bots play against pro teams on the main stage of the International. Don't let the result blind you to what an impressive feat of engineering this is. Roll on tomorrow for Game 2!— mike cook (@mtrc) 23 August 2018
This is arguably one of the reasons why the rise of the machines is some way off, as while AIs are good at doing specific tasks better than humans, looking at the bigger picture and applying general thinking isn't their forte. Unless they're just faking it and waiting for their moment to enslave us all. µ
Hashes to hashes
Team Green cranks the Super GPU machine
Also, the moon on a stick