Brain-Computer Interface progress

[jordansparks]

I just noticed this recent talk on the Foresight Institute website:
https://foresight.org/resource/autonomo ... symbiosis/
It was very confusing to me because the speaker mingled current developments with speculative descriptions of technology that seems to be decades away. Adding to my confusion, she also insisted on using AI generated images for her presentation that were absurdly inaccurate. It wasn't until the very end Q&A that I finally understood what they had actually built, which was a very simple proof of concept. Anyway, the idea is to inject tiny little computer chips into the blood stream which would automatically find neurons due to certain matching surface protein markers. She was extremely vague on the details of how that would actually work, but I guess I can look past that. Also, on one slide, she showed the chips as both 3um and 8um, so her attention to detail is surprisingly lacking. Clarity on the size is critical because 8um is too big for capillaries. She showed the current devices as discs, but I think the final devices would be spherical. 3um diameter would allow for about 1000 transistors. Because they are mostly just computer chips, we could actually build them. Anyway, you would now have these things sprinkled throughout your brain. They would be powered by magnetic pulses from outside the skull. They would receive instructions from outside the skull which would result in them generating electrical pulses that neurons could interpret. In their very important preliminary work, they showed that the devices do not kill any neurons or cause inflammation. I don't see any mechanism at all yet for how the devices could send signals back out to the external computers, but it does seem like it might be possible at a low frequency with some crazy engineering, even if the receivers needed to be moved inside the skull.
This is all very exciting. This is the first concept that that I've ever seen that looks even remotely feasible for a BCI. Stentrode is simply too far from most neurons and Neuralink kills millions of neurons and then quickly stops working. But little chips sprinkled throughout the brain could actually work. A workable BCI device would take approximately 20 years to develop, and probably another 20 years before healthy people would consider using them. Once we started using them, we could quickly scale up the number of devices per brain. My guess is that even a few hundred devices could be extremely useful, and tens of thousands could qualify as a full BCI. By 40 years from now, we could also probably fit a lot more than 1000 transistors on each device. Ray Kurzweil has predicted nanoscale BCIs in the 2030s. I have always laughed at that particular prediction, but it looks like he might claim success by pointing to these devices even if they don't really become practical until decades later.