Late on Tuesday evening, Elon Musk, the charismatic and eccentric CEO of SpaceX and Tesla, took to the stage at the California Academy of Sciences to make a big announcement. This time, he was not unveiling a new rocket or electric car but a system for recording the activity of thousands of neurons in the brain. With typical panache, Musk talked about putting this technology into a human brain by as early as next year.

The work is the product of Neuralink, a company Musk founded in 2016 to develop a high-bandwidth, implantable brain-computer interface (BCI). He says the initial goal is to enable people with quadriplegia to control a computer or smartphone using just their thoughts. But Musk’s vision is much more ambitious than that: he seeks to enable humans to “merge” with AI, giving people superhuman intelligence—an objective that is much more hype than an actual plan for new technology development.

Neuralink prototype device. Credit: Neuralink

On a more practical note, “the goal is to record from and stimulate [signals called] spikes in neurons” with an order of magnitude more bandwidth than what has been done to date and to have it be safe, Musk said at Tuesday’s event, which was livestreamed.

The system unveiled last night was a long way from Musk’s sci-fi vision. But it was nonetheless marked an impressive technical development. The team says it has now developed arrays with a very large number of “channels”—up to 3,072 flexible electrodes—which can be implanted in the brain’s outer layer, or cortex, using a surgical robot (a version of which was described as a “sewing machine” in a preprint paper posted on bioRxiv earlier this year). The electrodes are packaged in a small, implantable device containing custom-built integrated circuits, which connects to a USB port outside the brain (the team hopes to ultimately make the port wireless). Neuralink also intends to have the electrodes write signals back into the brain to provide sensory feedback in the form of touch or of visual stimulation of the retina in a blind person.* The company reported some initial results of its neural interface in rats in a white paper it made public, and it is currently doing experiments in monkeys at the University of California, Davis. None of this research has been peer-reviewed.

“More work in this area is great, and I think it’s fantastic that they’re giving this attention,” says Ken Shepard, a professor of electrical and biomedical engineering at Columbia University, who is part of a Defense Advanced Research Projects Agency initiative to develop a flexible, implantable wireless chip that uses electrodes on the surface of the brain to record up to a million neurons. Neuralink is focused on three themes that will be important to any future brain-computer interface technology, Shepard says: flexible materials for the electrodes, miniaturization of the electronics with integrated circuit technology and fully wireless interaction with outside devices. “They have made significant progress in the first two,” he says. But he adds that the challenges are going to be shrinking the electrical connections between the integrated circuits and the probes and incorporating many more electrodes without significantly increasing the size of the device. “The other big challenge is regulatory,” he says, noting the use of penetrating electrodes of this scale in humans is going to face significant hurdles from the U.S. Food and Drug Administration.

Neuralink surgical robot. Credit: Neuralink​

One of the big problems with existing electrodes is that they can damage vasculature when the brain moves as it does with each breath and heartbeat. The new device aims to get around this problem by using a small but rigid needle that inserts the flexible, polymer-based electrode “threads”—each a tenth the width of a human hair—into the cortex, taking care to avoid veins or arteries on the way in.

Perhaps the gold standard in neural recording for BCI research is the Utah Array, which consists of a rigid grid of up to 128 electrode channels. This array has been successfully used in a number of BCIs, including the BrainGate device developed by researchers at Brown University and their colleagues. But it also causes a tissue response that can lead to scarring of tissue composed of glia (the brain’s support cells), which may interfere with the quality of recorded signals or cause damage to brain cells. Another successful design is the Neuropixel, a probe developed by researchers at the Howard Hughes Medical Institute and their colleagues that consists of nearly 1,000 recording sites on a single tip, or shank, which can record from more than 500 neurons in the brains of mice. Developing tools for such high-density neural recording was one of the goals of the Obama administration’s BRAIN Initiative.

Leigh Hochberg, a professor at Brown University and one of the leaders of the BrainGate team, calls the Neuralink system “a novel and exciting” neurotechnology. “Given the great potential that intracortical brain-computer interfaces have to restore neurologic function for people with spinal cord injury, stroke, [amyotrophic lateral sclerosis], traumatic brain injury, or other diseases or injuries of the nervous system, I’m excited to see how [the company will] be translating [its] system toward initial clinical studies,” adds Hochberg, who is also a neurologist at Massachusetts General Hospital and the Providence VA Medical Center.

Neuralink claims its system can record from about 1,500 or 3,000 electrodes, depending on the version of the device being tested. The company asserts that because its electrodes are much thinner and more flexible, they are less likely to cause tissue damage. At the event, Musk said the reason for going with an invasive BCI—rather than one that detects neural signals outside the brain, such as electroencephalography—is that the company wants to record signals from individual neurons. “Everything we see, perceive or think are action potentials, or spikes,” he said. Musk noted that the ultimate goal is to make Neuralink’s device available to anyone, not just those with serious neurological illnesses, and to have it implanted in a minimally invasive procedure akin to LASIK eye surgery—though experts say such an achievement is a long way off.

The company hopes to begin its first human trial next year, the team said last night. This is an extremely ambitious target, given it still needs to obtain the necessary approval from the FDA, however.

Others in the field were gratified to get some transparency from a company that has shrouded itself in secrecy over the past few years. “Everyone is really appreciative that Elon has thrown his weight behind BCIs and brought visibility to the field,” says Matt Angle, founder and CEO of Paradromics—a firm that is also developing high-data-rate brain-computer interfaces—who is also part of the DARPA project. Angle is not as surprised by the technical developments, saying the achievements build on previous work by Neuralink senior scientist Philip “Flip” Sabes and bioengineer Timothy Hanson while they were both at the University of California, San Francisco. He notes that a challenge of such polymer-based electrodes is that they do not last as long as other inorganic materials when subjected to harsh conditions of the body. But he thinks Neuralink’s microfabrication expert Vanessa Tolosa and her team have made some impressive progress in materials science. Overall, Angle says, “I saw the most significant part of the announcement being [a willingness to] open up and engage with the community.”

*Editor’s Note (7/18/19): This sentence was edited after posting to clarify that the visual feedback would come in the form of electrical stimulation of the cortex, not the retina.