AI trained to slay players in a computer game could one day lead to killer robots in the real world, warn experts

  • The controversial software allows the AI to kill human players' avatars
  • The AI has been trained to play classic first-person shooter game Doom
  • Critics say it blurs the boundaries of what an AI should be trained to do

Two students have built an AI that could be the basis of future killer robots.

In a controversial move, the pair trained an AI bot to kill human players within the classic video game Doom.

Critics have expressed concern over the AI technology and the risk it could pose to humans in future.

Scroll down for video 

In a controversial move, two students trained an Artificial Intelligence bot to kill human players within the classic video game Doom

In a controversial move, two students trained an Artificial Intelligence bot to kill human players within the classic video game Doom

Devendra Chaplot and Guillaume Lample, from Carnegie Mellon University in Pittsburgh trained an AI bot - nicknamed Arnold - using 'deep reinforcement learning' techniques.

While Google's AI software had previously been shown to tackle vintage 2D Atari games such as Space Invaders, the students wanted to expand the technology to tackle three-dimensional first-person shooter games like Doom. 

Although other teams also developed similar technology to tackle Doom, the two Carnegie Mellon students published a paper online detailing their project, which has yet to peer-reviewed. 

Just like human players, the AI played the game repeatedly until it learned how to shoot enemies - including computer-based characters and human players' onscreen avatars. 

'Typically, deep reinforcement learning methods only utilise visual input for training,' explained the students in their paper.

'We present a method to augment these models to exploit game feature information such as the presence of enemies or items, during the training phase'. 

'Our architecture is also modularised to allow different models to be independently trained for different phases of the game.'  

Just like human players, the AI played Doom (pictured) repeatedly until it learned how to shoot enemies - including computer-based characters and human players' onscreen avatars

Just like human players, the AI played Doom (pictured) repeatedly until it learned how to shoot enemies - including computer-based characters and human players' onscreen avatars

Not only can the AI play the game, it is able to beat its human counterparts during multiplayer face-offs.

'We show that the proposed architecture substantially outperforms built-in AI agents of the game as well as humans in deathmatch scenarios,' said the authors of the paper.

Although the technology is undoubtedly impressive, critics have expressed concern over effectively training an AI to kill humans.

While there's no suggestion that the students' Arnold AI could or would ever be used in a scenario that might put humans in danger, it certainly raises important concerns over what could be possible. 

'The AI is as real as it gets. While it may have only been operating inside an environment of pixels, it does raise up questions about AI development in the real world,' said Dom Galeon, writing for Futurism.

'While we do not want to fall into the hype of AI hysteria, the importance of developing clear and sound policies about AI research and development and its applications are still to be considered,' he added.

While there's no suggestion that the students' Arnold AI could or would ever be used in a scenario that might put humans in danger, such as creating a killing machine like the The Terminator (pictured), it certainly raises important concerns over what could be possible

While there's no suggestion that the students' Arnold AI could or would ever be used in a scenario that might put humans in danger, such as creating a killing machine like the The Terminator (pictured), it certainly raises important concerns over what could be possible

In April, a report from the Human Rights Watch and the Harvard Law School International Human Rights Clinic called for a ban on 'killer robots'. 

The report called for humans to remain in control over all weapons systems at a time of rapid technological advances.

Last year, more than 1,000 technology and robotics experts — including scientist Stephen Hawking, Tesla Motors and SpaceX CEO Elon Musk and Apple co-founder Steve Wozniak — warned that AI weapons could be developed within years.

In an open letter, they argued that if any major military power pushes ahead with development of autonomous weapons and robotic cyber soldiers, 'a global arms race is virtually inevitable, and the endpoint of this technological trajectory is obvious: autonomous weapons will become the Kalashnikovs of tomorrow.' 

IT'S A 'NEAR CERTAINTY' TECHNOLOGY WILL THREATEN MAN  

It is a 'near certainty' that a major technological disaster will threaten humanity in the next 1,000 to 10,000 years.

That's according to physicist Stephen Hawking who claims science will likely bring about 'new ways things can go wrong' for human survival.

But the University of Cambridge professor added that a disaster on Earth will not spell the end of humanity – as long as humans find a way to spread out into space.

Hawking made the comments while recording the BBC's annual Reith Lectures on January 7.

The lecture explore research into black holes, and his warning was made during questions fielded by audience members.

When asked how the world will end, Hawking said that increasingly, most of the threats humanity faces come from progress in technology.

The scientist, who turned 74 this month, said these include nuclear war, catastrophic global warming and genetically engineered viruses.

The comments below have not been moderated.

The views expressed in the contents above are those of our users and do not necessarily reflect the views of MailOnline.

We are no longer accepting comments on this article.

Who is this week's top commenter? Find out now