In 2016 DARPA announced a program aiming at creating technologies to capture the brain’s activity at high resolution, meaning that one could look at activities of single neurones. The first target is to “read” up to a million neurones in parallel and researchers are confident that will be achieved at the end of this decade or in the first years of the next one. The challenges are not trivial and the result will still remain far from a real monitoring of a whole brain (even monitoring 1 million neurones one would be monitoring 1/100,000 of the brain! Actually it is even worse. The activity goes on at synaptic level getting the neurone level vision means you a one mile high vision -there are close to 1,000 synapses in each neurones…).
One of the problem is the transfer of the detected electrical activity to a computer. One might consider this as a non-problem given the capabilities we have achieved to transfer Gbits of data. Actually the issue is quite complex. First of all what we do with our computers is the transfer of “digital” data, bits, 0s and 1s. Neuronal activities creates spikes (voltage levels) that are basically analogous. These have to be translated into bits (in the same way that our analogous voice or an analogues stream of rays is translated into a sequence of bits that can be used to reconstruct the original signal with the desired fidelity. This “conversion” process is expensive in terms of power. Then the digitalises signal has to be sent to the web (to a computer somewhere) for processing and storage. And again, this requires power.
The amount of power required may be insignificant if you compare it to the power you use every day in most of your activity. However it becomes significant if you look at it in terms of heat dissipation and notice that such dissipation will have to take place “inside” your brain. I guess you are not looking forward to some implants cooking your neurones!
Here is where the results presented by a team of researchers at Lund University in Sweden. They have devised a very efficient processing architecture for electrophysiological set up that is capable of processing millions of spikes signals, translating each of them directly into a bit code, providing feedback within 25 millisecond, thus allowing a communication with neurones in the time window usable for communicating with them.
This result can foster clinical applications of brain computer interfaces.
It is interesting to see how different research groups all over the world are working to build the very many different pieces of the quilt needed to extend our brain into the cyberspace. Their aims is mostly addressing monitoring and potentially helping in disabilities but once that will be achieved we will have the technology for seamless human brain augmentation. We still need to sort out its implication…