Simulated ‘mannequins’ have been taught how to play soccer (football) by an international team of machine learning scientists. Yes, you read that correctly.
Sidenote: I insist on using the term “football” instead of “soccer”.
The results, including the endearing videos of the AI mannequins taking their new-found skills onto the pitch, are published in the journal Science Robotics. Do yourself a favour and watch those videos. If you think the screenshot above is the visual representation of chaos, you’ll love the movies.
But before the AI grudge match of the century, some basic training was necessary. To get the three-dimensional artificial intelligence humanoids to play football, they first needed to be coached in some rudimentary locomotion.
The team of computer scientists led by Siqi Liu is based at the British artificial intelligence and computer programming company DeepMind which is owned by Google’s Alphabet Inc.
In their paper, the researchers outline a three-step machine learning framework designed to teach the virtual AI humanoids a wide range of basic skills and motor functions.
Video demonstrating the machine learning study. Credit: Liu et al., Sci. Robot. 7, eabo0235.
First, the AI mannequins watched videos of moving humans to learn through imitation to mimic the low-level natural movements such as walking.
Next, a reward-based reinforcement learning approach taught the AI mid-level skills such as kicking a ball.
Finally, they gained high-level skills, including teamwork and body control, by working together in teams under an advanced version of reward-based reinforcement learning. Also known as a “match”.
Read more: Artificial intelligence tool swings into action, discovering alternative physics variables!
No prior knowledge of football, its rules or its objectives were fed into the machine learning algorithm. Yet the virtual players learned all these skills and played a tensely fought game against each other exhibiting collaboration, body control and decent technique. And, as a casual futsal administrator of six years, I should know.
Much like taking a round ball into a rugby- or Aussie rules-loving town, the researchers provided no background to the game itself, yet the AI was able to learn how to play football through machine learning. Kicking goals indeed.
“The result is a team of coordinated humanoid football players that exhibit complex behaviour at different scales, quantified by a range of analysis and statistics, including those used in real-world sport analytics,” the authors write.
Among the familiar football skills mastered by the virtual players were jostling for position, clearing the ball away from the goal they are defending, rapid turns, through balls, kicks of varying height, slowing down before changing direction with the ball, tackling and running into a defensive position.
Lionel Messi and Cristiano Ronaldo they aren’t, but these simulated soccer stars would have walked right into my high school football team.
Apart from giving hilarious reprieve and emulating our proud A-League, what do the researchers hope to achieve?
“[The simulated players] achieved integrated control in a setting where movement skills and high-level goal-directed behaviour were tightly coupled – a setting that is reflective of many challenges faced by animals and humans and where solutions would be extremely difficult to handcraft,” the authors write.
They believe that research into teaching AI how to move and interact like animals or humans may help in the development of more animal-like robots.
“Enabling machines to produce agile, animal-like movements has been a goal of robotics research,” they write. “To date, these techniques appear to be rather distinct from the learning-based solutions developed in the AI community in simulation. Yet, recent partial successes in transferring results from simulation to real robots suggested that learning-based approaches in simulation may, in the future, play a larger role in the control of real-world robots.”
With the World Cup only a few months away, watching these clumsy AI mannequins going at it has certainly got me in the football spirit.
Evrim Yazgin has a Bachelor of Science majoring in mathematical physics and a Master of Science in physics, both from the University of Melbourne.
Read science facts, not fiction...
There’s never been a more important time to explain the facts, cherish evidence-based knowledge and to showcase the latest scientific, technological and engineering breakthroughs. Cosmos is published by The Royal Institution of Australia, a charity dedicated to connecting people with the world of science. Financial contributions, however big or small, help us provide access to trusted science information at a time when the world needs it most. Please support us by making a donation or purchasing a subscription today.