Elon Musk's Neuralink

Elon Musk recently presented his vision and the team's progress on creating an interface to the brain's neurons. The basic idea is to design a system which is able to read the activity of neurons in order to read what the brain is "thinking" at any given time. The simplest use of such an interface would to be allow users to, for example, think about moving forward, and then have a character in a game move forward. Such "natural" interfaces, would make it much easier to use computers, since instead of having to figure out how to do something, the user simply thinks about what he or she wants to do and the computer figures out the rest. Imagine a Google search where instead of typing "cute puppies", the user just thinks about wanting to see cute puppies and Google figures out the rest.

However, Elon was clear that this simple application is not where Neuralink wants to stop. More than just "mind reading", Neuralink will be designed to actually "write" electric impulses to the brain. Basically, instead of just measuring the so-called action potential of neurons, the system would also stimulate those neurons in order to cause them to fire. Theoretically, the effect would be to make you think certain things. For example, instead of Google displaying images of cute puppies in your web browser, it would make you visualize them by stimulating your neurons. The impact of such "mind control" is unclear, and when I heard that this system would be both read and write, I was highly skeptical.

Much like gene editing, I cannot imagine that even with artificially intelligent algorithms the impact of neural stimulation could be known. Rather, the effects might be like random shots in the dark that might have the desired impact, but then might instead (or additionally) cause some other effect. Google might try to make you see a puppy, but instead you could see a dead puppy or something worse, or even just fall right out of your chair. Although the team mentions that they intend to deliver an early stage of the product by next year, I'm sure that this product will just be the hardware necessary to perform such reads and writes, and won't actually implement any of the desired features.

As with all things biological, trying to understand what is going on is almost always an uphill battle. Evolution certainly designed very powerful systems, but the way that it designed them means that they are incredibly complex and nuanced. They are nothing like the systems that engineers design, and unfortunately engineers have a long history of underestimating that complexity. In the 60s, for example, there was work done to try and emulate the brain. Cognitive scientists hoped that if they could understand how the brain worked, they could just design that same functionality with a computer and thus create an artificial intelligence. However, the entire field crumbled when it was discovered just how much coupling there is in the human's body. For example, the idea that our body sends the brain raw signals through our eyes into the brain is completely false. Instead, a lot of the visual processing is already done in the eyes themselves. This coupling of components, something that computer scientists work very hard to avoid, creates systems which are very complicated to understand, because it's never clear where the signals are coming from or what they mean.

Elon Musk and Neuralink are bold to think that they can tackle this problem, and no doubt a lot of progress will be made. Unfortunately, it's unclear what the cost might be of mistakes made in this realm. Stimulating the brain in the wrong way might turn users into vegetables, or else turn them into psychopaths. Honestly, as silly as it all sounds, it might happen. The point is: no one knows.