M3GAN: The strange reasons a real child AI robot could never be built
For good or bad, a kid's toy with artificial intelligence (and killer dance moves) probably won't be possible at any point soon.
New horror movie M3GAN is essentially Chucky meets the Terminator. Its titular star is posited as the future of mass-market robots: a lifelike doll that has been programmed to be a child’s best friend.
It can listen, it can learn, it can dance (trust me, watch the trailer, it can dance) and it can also, in what is surely a big design flaw, kill. Not ideal. Deadly robots are nothing new, of course, but how feasible is the idea that a non-homicidal version of M3GAN could appear on shop shelves anytime soon?
To answer that question, we must first examine what M3GAN is; an apparent marvel of artificial intelligence (AI). “An autonomous robot does think for itself, but the scope of its actions are very limited,” says Prof Peter Bentley, a computer scientist based at University College London and a regular contributor to BBC Science Focus.
“More often than not, even the most advanced AIs are just big encyclopedias of our nonsense. They’re amazing at what they can do, but there’s no consciousness there. There’s no understanding. There’s certainly no emotion. We’re very far away from anything remotely close to M3GAN.”
And that, according to Bentley, also includes the idea of an AI capable of autonomous violence. “How do you tell an AI never to kill, except when we’re at war?” he asks.
“What if a man is holding a gun to a child’s head? Should the AI shoot? What if it’s a toy gun and the man is the child’s father and the robot blows his head off? Building in morality and experience of the world is not something we’ve got the hang of.”
Indeed, the biggest danger to us is not ‘evil’ AI, he says, “but that we believe AI is better than it actually is. Just look at self-driving cars.”
Why we'll never have a robot that dances like M3GAN
M3GAN is not only a fantastical vision of artificial intelligence, she is also a genius feat of robotics. “With current humanoid robots, you’re lucky if they don’t fall over most of the time,” says Bentley.
“We can’t make a robot as complex as the human body. And even if we could, we don’t have a power source that would work. There are a lot of actuators to think about: all the different joints, movements and muscles we’ve got.”
In reality, says Bentley, the robot would either require an “enormous battery backpack” or would have to be plugged into a wall permanently.
As for M3GAN’s souped-up TikTok dance moves, Bentley doubts that robots could ever achieve such sass. “In robotics, we talk about degrees of freedom [joints] and a robot arm might not have that many,” he says.
“It might just have an elbow, a wrist, some at the hands. Maybe four or six degrees of freedom. But that dance requires a level of precision that is currently not possible.” This is why robots are famous for their jerky dancing. “Robots don’t have sufficient degrees of freedom to achieve that fluidity of movement.”
Yet Bentley’s main argument against a doll like M3GAN has more to do with business than technology. He points out that recent advances in the field of AI have been driven by companies such as Google and Meta (Facebook), because there is a “clear business case” for an AI that can recognise cars or faces or fingerprints.
“But to develop a robot as sophisticated as M3GAN would cost billions, maybe trillions, of dollars,” he says. “Is there a good business case for a freaky little humanoid robot that might kill us? It would be way more complex than your autonomous family car and therefore the price tag is going to be equal, if not more [than a self-driving car]. Ultimately, the best example of a humanoid child-shaped intelligent friend for your child is another child.”
Thank goodness, creepy AI robot dolls will not be appearing in any toy shops. For now…
Read more about AI:
Stephen Kelly is a freelance culture and science journalist. He oversees BBC Science Focus's Popcorn Science feature, where every month we get an expert to weigh in on the plausibility of a newly released TV show or film. Beyond BBC Science Focus, he has written for such publications as The Guardian, The Telegraph, The I, BBC Culture, Wired, Total Film, Radio Times and Entertainment Weekly. He is a big fan of Studio Ghibli movies, the apparent football team Tottenham Hotspur and writing short biographies in the third person.