by Lisa Jarvis
Two new studies published simultaneously in the journal Nature document a leap forward in the race to teach computers to translate brain signals into text. It’s an exciting development in a field that is attracting millions in investment, including to Elon Musk’s brain-implant company, Neuralink Corp.
To turn these discoveries into viable commercial products, investors and entrepreneurs should focus less on Musk’s musings about a world where able-bodied people merge with computers to enhance their intelligence. They should instead focus on the monumental benefits of helping people who have lost their ability to communicate or move.
Advances in technology are bringing researchers tantalizingly close to restoring the ability to speak to those who have lost it, whether because of stroke, paralysis or a disease like ALS. Crucially, that work holds lessons that could accelerate broader efforts to connect our brains to computers, including those intended to help paralyzed people regain control over their muscles.
In the Nature studies, two teams, one led by researchers from Stanford University and the other led by University of California San Francisco, used different types of implants to collect electronic signals from the brains of volunteers and different algorithms to interpret them. The results, though, were similar: Attempts at speech could be converted to text at a rate of 60-70 words per minute. While still half as fast as a typical person’s cadence, it’s a vast improvement over earlier brain-computer interfaces.
And in both studies, the computer was able to decode speech with remarkable accuracy. Whereas earlier versions of the technology could decipher only a few hundred words, the Stanford team’s algorithm could interpret the equivalent of a short dictionary.
In addition to improving the range and speed at which someone can speak, the UCSF team took the user experience a step further: They developed a personalized avatar for their volunteer, who was paralyzed by a brainstem stroke more than 18 years ago. The avatar uses her own voice, taken from a clip of her speaking at her wedding, and integrates facial expressions. “Speech isn’t just about communicating words, but also who we are,” says Edward Chang, the neurosurgeon who led the UCSF work. “Our voices are our identity.”
Although each team reported results from just one volunteer, other researchers in the field say it’s enough to be confident that brain-computer interfaces (BCI) are nearing a point where they could be truly useful to people unable to speak. An accompanying editorial by two other brain-machine experts called the developments “a turning point in the development of BCI technology that aims to restore communication for people with severe paralysis.”
All this hope must be grounded in an unfortunately harsh reality: It will be very tough to turn these technologies into profitable products. The eligible patient population is small. The devices will be expensive and require very specific expertise and training to use. Early entrants to the market will face a hard path. The field’s most prominent cautionary tale is the meltdown of Second Sight Medical Products, a company that developed retinal implants. In 2020, the company began winding down operations before eventually merging with another firm, leaving people who relied on their bionic eyes in the lurch.
Yet eventually, investment in this field is likely to pay off as researchers find broader applications. “As soon as you have the technology to read from a large number of neurons, or stimulate a large number of neurons, these technologies will be used for many things,” says Cynthia Chestek, a biomedical engineer at the University of Michigan. Chestek, for example, can imagine how similar hardware and algorithms could help with her own work trying to use these types of signals to read muscle commands in people with spinal cord injuries — and in turn help them move prosthetic or even their own limbs.
Too much of the hype around BCI technology has centered on Musk’s musings about a world where cyborg-humans operate with enhanced consciousness. To be clear: If ever achievable, that’s a dream that’s many decades in the future.
Where Musk has helped is by stimulating interest and investment in the field. That should help get the hardware (the actual brain implant) to a point that regulators feel comfortable with its long-term safety.
On a call with media, Stanford researcher Jaimie Henderson reminded reporters of what’s at stake for families. When he was 5, his father lost much of his ability to communicate after a devastating car accident. As a child, he could sometimes understand the joke his dad was trying to tell, but not the punchline. “I grew up wishing that I could know him and communicate with him.”
It’s not a stretch to imagine the feeling was mutual. These advances suggest Henderson’s wish could be a reality for many families in the not-to-distant future.
