One of the biggest areas of VC funding lately seems to be brain computer interfaces. There are so many firms working on it, with the most famous being Elon Musk’s Neuralink, which claims to have had some success with connecting electrodes to the brain of a monkey. The technology has promise to make technology of the future like brain operated cars and vehicles possible and to revolutionize prosthetics, and it is undeniably cool.
But it has some issues. First among which is that the brain really doesn’t like sharp metal bits inside the skull, and so no solution has yet survived for longer than a month. This is a more intractable problem than it might sound like. The brain is a complex enviroment, and the most well protected system in the body. There’s so much going on that electrodes have a hard time maintaining reliability in such a hostile system. That’s ignoring the fact that within weeks, any metal is covered in a protective organic layer to protect the brain from damage, ruining any connection. So far there is no drug that can overcome this. Therefore I think that these interfaces will take a while to build, much longer than a decade or a few, since its interfacing with such a fundamental system.
When they interface though even, what they’ll do is also questionable. Beyond prosthetics, which is a good field in and of itself, there is the question of whether given the recent advancements in AI control and management, that whether any control mechanism you could think of using the brain couldn’t be better approximated in the same time frame by an AI assistant which would be quicker and more autonomous. Of course there is the even further future idea of a AI assistant and your brain operating in concert to immediately make changes, but whether it would be clearer and easier than just simply using text or video is questionable, and of course it would raise the issue of adverse thoughts messing up control, which is not an issue usually with more restricted input mechanisms.
Feeding information into the mind too is also interesting. What would the interface even look like? How could it interface? The brain is not a simple program, all the information is stored by subtle growths on the neurons, and so information creation, transfer and control would all be very difficult, and of questionable utility given the danger involved. It probably will eventually be solved, but these are formidable challenges before any sort of system like this. That’s why I think that the market should be more bearish on these systems. In 2084, there will be some, but not as much as AI, and interfacing will still look to some extent, somewhat like today - text and video based, as that’s easy to clean, manage and interpret, and removed from the tricky issues of modifying the brain.
The interface may have to be organic. The result of the application of quadrillion parameter neural network intelligences to the problem. The beauty of these new neural networks is how it integrates knowledge - shown beautifully in the way natural language and images can flow together to form a novel image built from the subtle and simultaneous contribution all images and language ingested. So I think that the problem of the neural interface is hard because we have to limit our ambition due to the lack of our ability to integrate all of the possibilities in a novel new solution. The AI will build its own interface to us.