exciting! you could imagine brain transcriptions being very high signal RLHF data that could help align models more deeply to human preference.

are you specifically focused on non-invasive methods/why? seems like you reach a noise barrier which limits things like this