This will be very useful for brain researchers (and scarily interrogators). I wonder if you could do a similar thing and use an LLM to extract text/language from MRI imaging to read what a person is actually thinking.