MISSION IMPOSSIBLE? Elon Musk’s quest to stop the ‘AI apocalypse’
Musk makes good on his concerns over risk of artificial intelligence
● If you can’t beat them, join them. Or if you’re Elon Musk, the billionaire boss of private rocket outfit SpaceX and electric vehicle firm Tesla, why not just become them?
The tech mogul unveiled plans this week to implant human brains with computer chips in an effort to merge man with machine, the start of a quest to prevent Silicon Valley’s most feared scenario: society’s demise at the hands of artificial intelligence.
Musk has long sounded the alarm on AI, claiming that rapid advances in machine learning risk “summoning a demon” that leaves humans in the dust as super-intelligent systems outmanoeuvre our species.
“At least when there’s an evil dictator, that human is going to die. But for an AI, there would be no death. You’d have an immortal dictator from which we can never escape,” he has warned.
Musk told an audience at the California Academy of Sciences that Neuralink, the private vehicle for this vision, has been testing its “brain-computer interface” on monkeys and rats, and is expected to begin trials on humans by the end of 2020.
The prize at hand is clear, with Neuralink setting its sights on not just competing with AI by accelerating humanity’s path towards a future of super-intelligence, but also on solving complex neurological disorders.
Investors are starting to support the vision, with Neuralink, whose technology has barely been tested, raising $150m (R2.1bn) since its founding in 2016, including $100m from Musk himself. It has also attracted some of the brightest minds in science.
But others are less convinced. Even for Musk, the task of solving brain disorders and advancing human intelligence is a tall order.
The first hurdle is likely to be a regulatory one. Daniel Mansson, a clinical psychologist, has been working on a “brain stimulation technique” to treat depression with his medical device startup, Flow, and has seen how long the regulatory process can take for something that is non-invasive.
Neuralink’s technology is more complex and intrusive. The startup has devised a 4mmx4mm chip that connects to a thousand microscopic threads entering the brain through four holes drilled in the skull.
The threads, tagged with electrodes, could theoretically detect electrical impulses in the brain to track activity in the body’s primary
An evil dictator, that human is going to die. But for AI, there’d be no death Elon Musk
SA-born technology entrepreneur
nerve centre, which would then feed back as data on a smartphone. The prospect of human trials by 2020 is likely to be a stretch when the US Food and Drug Administration stands in the way.
The challenge has been acknowledged by Neuralink’s president, Max Hodak, who said the company was “under no illusion” about achieving its goal alone and would require help from others. It is unclear, too, how Musk plans on using the chips to enhance human intelligence and solve brain disorders, with no real data presented as yet.
“I think the possibilities for application of the Neuralink technology are pretty limited — it’s not going to make us smarter,” says David Curtis, a specialist in genetics and psychiatry at University College London.
What is more plausible, Curtis says, is an application in which the brain implants allow patients with things such as locked-in syndrome to “communicate with the outside world”, and possibly for people with motor neuron disease to gain better control of artificial body parts.
A primary driver for Musk has been the “democratisation of intelligence” to ensure humanity is on equal pegging with AI, but not everyone is convinced of a doomsday case — or even the idea that humans and AI will be in direct opposition.
“Fears about an AI takeover are not justified by any research or evidence … the idea of implanting stuff in our brains to keep up with AI is just nonsense,” says Noel Sharkey, emeritus professor of AI and robotics at the University of Sheffield.
Alan Turing, the mathematician widely considered a founding father of AI, suggested almost 70 years ago that there is no known reason a computer can’t do everything a human brain does. But he also suggested that if machines did eventually compete with people, where would they start?
It is for this reason that Musk’s dreams of a “symbiosis” with AI may not be necessary and remain, for now, just a distant possibility.