It feels kind of ridiculous to say this, but Elon Musk’s most ambitious project isn’t the one dedicated to sending a rocket ship full of people to Mars. Nope.
That’s a breeze compared to Neuralink, the biotech startup he founded in 2016, which is trying to link the human brain with computers.
On Friday, Musk was joined by a cadre of scientists, surgeons and animal technicians from Neuralink for a livestreamed “progress update” of its latest brain-computer interface, officially known as the Link. “In a lot of ways, it’s kind of like a Fitbit in your skull with tiny wires,” Musk proclaimed.
The presentation also featured the company’s newest version of its “sewing machine” surgical robot, which can implant thin wires into the brain, and three pigs. One, the audience was told, had a “working” Neuralink implant in its skull.
Musk stated from the outset that the “main focus” of the event was recruiting talent to work at Neuralink, but that didn’t prevent him from delivering a spectacle, of sorts. The intriguing presentation — featuring a flashy Hollywood trailer and disobedient pigs — was capped off by an in-depth Q&A session that dealt with more technical aspects of Neuralink’s plan to replace a small section of your skull with a dollar-sized gizmo that can read your mind.
Musk spent a lot of time detailing the Link and its possibilities, but was scant on specifics and provided no hard data on experiments performed by the company. Instead, he often mused about a sci-fi future where we might play back memories, achieve symbiosis with artificial intelligence or give people super vision allowing them to see in wavelengths of light the human eye cannot normally see. Other employees remarked on exploring the nature of consciousness itself.
These grandiose imagined futures contrast dramatically with the reality of Neuralink’s current progress. Yes, the future of the tech could certainly be life-changing, but the lightning-fast presentation underscored how much Neuralink has focused on the “computer” in “brain-computer interface.” Since the launch conference in 2019, Musk’s startup has made significant improvements to its device and overcome significant technical challenges that could hamper implantation.
But it doesn’t seem as much time has been spent thinking about the “brain.”
That’ll do, pig
As is always the case with Musk’s big, live reveals of his works in progress — he’s done it with SpaceX, he’s done it with Tesla — there was a keen interest in what he had to show off, driven by hype-building tweets he let fly to his 38 million followers. Hundreds of thousands tuned into the livestream, and millions have since watched the presentation on YouTube.
The stars of the update were a trio of experimental pigs. During the stream, Musk literally pulled back a curtain to reveal the swine, which had each contributed to testing Neuralink’s latest device. Musk explained one little piggy had no Link, one little piggy had one (but it had been removed) and the third still bore the implant.
Gertrude, pig No. 3, shuffled its snout about a small pen, and viewers were able to hear and see what Musk described as real-time signals of its brain cells firing on a readout that flashed up on screen. The Neuralink in Gertrude’s head, according to Musk, was connected to cells in her snout. Every time Gertrude touched her snout to something, the signals spiked, beeping and blooping for the audience.
In a subsequent part of the presentation, Musk showed off a short prerecorded video of a pig, presumably wearing a Link device, walking on a treadmill. The device was reading the pig’s brain activity as it walked.
“We take the readings from the neurons, and we try to predict the position of the joints,” Musk explained, declaring that the Neuralink device could do so “with very high accuracy.” However, the video appeared to show significant differences in the predicted position and the actual position. That’s not to say Musk is lying about the results or that Neuralink concocted a misleading presentation, it just serves to highlight how difficult the brain really is to decode.
It’s unclear how big these differences really are right now. If the Neuralink was controlling the pig’s movement, would it completely fall down? Would it break a leg? What would its gait be like? We simply don’t know because we haven’t been able to see any data. We don’t know where the thin wires of the Neuralink connect to the pigs’ brain, how durable they might be, how long they can last implanted in the brain or what the signals we’ve seen really mean.
Neuralink didn’t respond to a request for comment.
Musk and Neuralink may yet release raw data related to Friday’s update in a peer-reviewed journal, like they eventually did after the Neuralink launch conference in 2019. But Musk also made huge, as-yet-unverified claims about a monkey “able to control a computer with its brain” at that same conference and we’re yet to see those results. Strangely, the Neuralink website was updated following the progress update, removing links to the summer launch event and the peer-reviewed paper from October 2019.
Angus McMorland, a bioengineer at the University of Auckland, New Zealand, said there’s nothing contentious about the theory behind Neuralink’s work and notes he isn’t too concerned about the data, but suggests it would be nice to know details about the quality and stability of the units.
The complexity of the human brain is staggering. Almost incomprehensible.
“In the brain, we have some 80 billion neurons, and many times that number of connections between them,” says McMorland.
All of these connections, which bounce electrical impulses around your brain, are responsible for you. Your thoughts, movements, speech, beliefs, social interactions… your consciousness. You are your brain cells, firing away. You are your brain cells, reading this article.
But experts who study the brain only have a rudimentary knowledge of how these connections, and their location in your brain, are intertwined and control functions. Neuroscientists can point to patches in the brain that light up when you’re scrolling down this page, but they can’t tell you what is controlling that action. While they are getting better at pinpointing which regions and impulses respond to what, we simply don’t have a good “map” of the brain.
What Neuralink may be able to do is pioneer a better way to read brain signals, accelerating the pace of research in the area. But understanding those signals is inherently more difficult. Manipulating them could be even harder.
McMorland offers a good example. During the presentation, it was suggested you might one day be able to summon a Tesla just by thinking about it. But the act of summoning a Tesla may differ each time you do it. Different parts of the brain and unique patterns of brain activity might be required, depending on the situation.
“That pattern could vary depending on context: Are you in a restaurant, in a hurry, standing by the side of the road or performing a bank robbery?” asks McMorland.
Does it really matter right now, four years after Neuralink was founded? Perhaps not to Musk.
As McMorland rightly points out, this approach is very similar to what the tech billionaire has done with SpaceX and Tesla — take existing technologies, tinker away, make small improvements and package them up in high-spec, neatly wrapped products.
Can this roadmap work for Neuralink, too?
Potentially, but deciphering and decoding the brain is a mammoth problem that will need to be solved before any of the fanciful applications touted by Musk and his colleagues can come to fruition. And it’s a problem that is magnitudes greater than building rockets and electric vehicles. The technology may wow us, but you can’t put the cart before the horse — or the pig, as it were.