Neuralink, the Elon Musk-owned brain-computer interface company, has received “breakthrough device” clearance from the FDA. But this does not mean they have developed a cure for blindness, whatever Musk might say.
The breakthrough devices program at the FDA is a voluntary program that developers can apply to that, if granted, “offers manufacturers an opportunity to interact with FDA experts through several different program options to efficiently address topics as they arise during the premarket review phase,” as well as priority for FDA review.
In 2023 more than 30 devices were designated this way, and almost 1,000 have been authorized total. Neuralink’s Blindsight is the latest, as the company announced Tuesday on X. Musk lost no time in asserting that it will “enable even those who have lost both eyes and their optic nerve to see… it will even enable those who have been blind from birth to see for the first time.” That’s not quite true.
Blindsight is a new iteration of a technology that has been used for decades to experimentally restore very limited vision to certain blind people. A microelectrode array is embedded in the visual cortex, and stimulates the neurons there in patterns derived from a camera. In a way, it really is that simple to produce visual phenomena in people who may indeed never have seen before.
But it is at best extremely premature to say that such a device could enable blind people to see. The issue historically has been the low density of electrodes on the array: on the order of dozens, meaning what is “seen” is really more like a few stars winking on and off with no discernible pattern, because the parts of the cortex pierced and stimulated are essentially random.
Neuralink’s advance — which is very welcome in this field — is increasing that density. But other than that, the approach suffers from the same fundamental drawbacks.
What Musk is implying is that the device would plug in and grant sight. That is not possible. Even among those who have only recently lost their sight to trauma or disease (by far the most common causes), meaning the visual cortex has been trained since birth to function normally, sight would not be “like Atari graphics,” as Musk suggests. While there is a surprising and promising level of plasticity in adapting to systems like this, it is a difficult and disorienting process.
Furthermore, people who have been blind from birth will not have developed the biological capacity for seeing through their eyes, meaning that despite the visual cortex’s cellular layout being optimized for vision tasks, the pathways that create the concept of vision sighted people understand will not exist. It is misleading for Musk to suggest otherwise, though I suspect the blind and low-vision community is used to sighted people making this kind of mistake.
This is not all to say that Neuralink’s Blindsight is bad or won’t work — the company appears to have created a genuinely better microelectrode array, and possibly the means to implant it more efficiently and with lower risk of rejection or brain damage.
While it’s irresponsible for Musk to make the same bombastic “Geordi La Forge” promises that others have made these last few decades in this area of research, the device and methods he has funded will almost certainly contribute towards a future where electronic implant-based treatments for blindness become realistic.
If Musk’s companies ever responded to press inquiries (Neuralink did not immediately reply to a request for comment) or discussed and published their research more openly, we would have a more open discussion among domain experts about the merits and challenges of this approach. As it is, we may hope that, now that the FDA has given them permission to take their shot at justifying the device’s use in humans, the company will operate more in the open than it has hitherto.
