The bit about single-participant models being unable to generalize is really interesting.
> there are no people who exhibit the ability to generate performant models for other people, nor are there any people for whom other people’s models always perform well
I wonder just how unique an individual's motor neural signaling is? If 100 participants was not enough to expose even a single broad overlap, could it really be so varied as to be near unique among every person? Or does the success of the larger corpus model suggest that the overlap is not being collected and/or exposed well by this methodology?
Blackrock Neurotech has some BCI cortical array chip-to-emg bridge, I think; one was implanted at Case Western a number of years ago. the EMG sensors are probably similar.
The bit about single-participant models being unable to generalize is really interesting.
> there are no people who exhibit the ability to generate performant models for other people, nor are there any people for whom other people’s models always perform well
I wonder just how unique an individual's motor neural signaling is? If 100 participants was not enough to expose even a single broad overlap, could it really be so varied as to be near unique among every person? Or does the success of the larger corpus model suggest that the overlap is not being collected and/or exposed well by this methodology?
The videos are great:
https://static-content.springer.com/esm/art%3A10.1038%2Fs415...
https://static-content.springer.com/esm/art%3A10.1038%2Fs415...
https://static-content.springer.com/esm/art%3A10.1038%2Fs415...
Blackrock Neurotech has some BCI cortical array chip-to-emg bridge, I think; one was implanted at Case Western a number of years ago. the EMG sensors are probably similar.