Rise of the Machines: How an AI Gaming Bot Took On the Humans…And Won

It’s a classic trope of Hollywood: AI-powered robots and machines going rogue and taking over the world.

And while Tinseltown revels in tales of post-apocalyptic worlds dominated by our metallic overlords, the truth is that – for now at least – there’s no danger of society being run by deranged cyborgs hellbent on overcoming their human paymasters.

However, perhaps that theory took one miniscule step closer recently.

‘Quake III’ – BCAOE Loguidice via Flickr (CC BY-SA 2.0)

Surely when the uprising comes, it will emanate from Google’s DeepMind artificial intelligence lab, which features a collection of the sharpest minds in AI development.

Their latest success was to build an algorithm that was able to take on human opponents in the classic multiplayer game Quake III Arena…and win.

Yes, that’s right: a piece of AI software was able to outthink and outmanoeuvre human competitors in the game’s legendary Capture the Flag mode.

Are you getting scared yet?

Okay, so this is hardly a development that foretells the end of humankind as we know it. But it is another astonishing leap forward in AI programming, to go with similar case studies of artificial intelligence defeating two of the best StarCraft players in the world at that particular title.

But what made this experiment so intriguing is that Capture the Flag requires teamwork; this is a multiplayer game where a roster can be split into sections: one responsible for capturing their flag and the other tasked with repelling enemies and protecting the flag bearer by shooting anything that moves.

And, rather alarmingly, the AI bots seemed to work like a, well, a well-oiled machine.

The “For the Win” outfit, as the AI team was known, was allowed 12 hours of machine learning by practicing the various game maps before they were set loose against their humanoid opponents.

In all of the test games played, the AI-based unit won 75% of the battles fought.

Amazingly, the teams were then switched up to pair an AI bot with a human player, and that gave the duo a 5% greater win probability as the CPU was able to work seamlessly and “read” the movements of its human companion.

A Timeline of Staggering Achievement

So what does the DeepMind project tell us? Perhaps not all that much, other than an AI bot can be taught to play an enjoyable first-person shooter epic rather well.

But it was the manner in which a squad of AI-based units were able to work together – and subsequently learn and aid the human player – that is the most interesting output of the experiment.

It continues the incredible forward leap in artificial intelligence that has occurred since the days of Alan Turing and the test named after himself, which queried whether a robot could convince people it was human.

Source: The BBC’s timeline of AI reproduced by A2Zcasinos

The timeline shown above follows the journey from fledgling technology to something that now pervades everyday life, with artificial programming used in healthcare, logistics and retail already and with many more use cases being developed all the time.

Soon, we will all be on a journey into a brighter future… and taken there in our self-driving cars, of course.