ELON MUSK, perhaps the world’s most famous entrepreneur, is sometimes referred to as “the Trump of technology”—not for political reasons, but because of his habit of making, at short notice, spectacular pronouncements that stretch the bounds of credibility. On July 16th he was at it again, unveiling a new type of brain-machine interface (BMI). If human beings do not enter a symbiosis with artificial intelligence (AI), he declared, they are sure to be left behind. And he, the announcement implied, was going to be the man who stopped that happening.
Connecting brains directly to machines is a long-standing aspiration. And it is already happening, albeit in a crude way. In deep-brain stimulation, for example, neurosurgeons implant a few electrodes into a patient’s brain in order to treat Parkinson’s disease. Utah arrays, collections of 100 conductive silicon needles, are now employed experimentally to record brain waves. A team at the University of Washington has built a “brain-to-brain network” that allows people to play games with each other using just their thoughts. And researchers at the University of California, San Francisco, have captured neural signals from people as they talk, and have then turned that information, via a computer, into intelligible speech.
As with all things Musk-related, Neuralink is much more ambitious. The firm does not just want to develop a better BMI. Its aim is to create a “neural lace”, a mesh of ultra-thin electrodes that capture as much information from the brain as possible. Unsurprisingly, hurdles abound. The electrodes needed to do this must be flexible, so that they do not damage brain tissue and will also last for a long time. They have to number at least in the thousands, to provide sufficient bandwidth. And to make the implantation of so many electrodes safe, painless and effective, the process has to be automated, much like LASIK surgery, which uses lasers to correct eyesight.
Neuralink does indeed seem to have made progress towards these goals. Its presentation, at the California Academy of Sciences, in San Francisco, included videos of a neurosurgical robot that is best described as a sewing machine. This robot grabs “threads” (films, containing electrodes, that measure less than a quarter of the diameter of a human hair), and shoots them deep into the brain through a hole in the skull. It is capable of inserting six threads, each carrying 32 electrodes, per minute. The firm has also designed a chip that can handle signals from as many as 3,072 electrodes—ten times more than the best current systems—and transmit them wirelessly.
The real magic, however, kicks in only when the output is analysed—which happens in real time. Looked at superficially, neurons in the brain seem to fire at random. Software can, though, detect patterns when the individual those neurons are in does certain things. Stick enough electrodes into someone’s motor cortex, for instance, and it is possible to record what happens in the brain when he types on a keyboard or moves a mouse around. Those data can then be used to control a computer directly. Conversely, the electrodes can be employed to stimulate neurons, perhaps to give the person in question the feeling of touching something.
Neuralink has already tested its system successfully on rats and monkeys. These were, it says, able to move cursors on screens with it. The firm now hopes to work with human volunteers, perhaps as early as next year should America’s Food and Drug Administration play along.
The first goal is to use the technology to help people overcome such ailments as blindness and paralysis. Neuralink is, however, clearly aiming for a bigger market than this. It has also designed a small device that would sit behind someone’s ear, picking up signals from the implanted chip and passing them on as appropriate. In a few years, using a brain implant to control your devices may be as de rigueur among San Francisco’s techno-chics as wearing wireless earbuds is today. Ultimately, Mr Musk predicts, neural lace will allow humans to merge with AI systems, thus enabling the species to survive.
Though, as this announcement shows, Mr Musk does have a habit of presenting himself as the saviour of the human race (his desire to settle Mars seems motivated partly by fear of what might, in the future, happen to Earth), the idea that some machines at least will come under the direct control of human brains seems plausible. The biggest obstruction to this happening will probably not be writing the software needed to interpret brainwaves, but rather persuading people that the necessary surgery, whether by sewing machine or otherwise, is actually a good idea.■
Correction (July 18th 2019): Neuralink’s new chip can handle signals from ten times as many electrodes as the best current systems, not 30 times as many, as we first wrote.