This year, we have worked on an alpha release of an AI-based chatbot to help analyze FMRI data. Here is an excerpt from a session with afniGPT*:
AFNI_user: Hello afniGPT
afniGPT: Hello
AFNI_user: How can I be sure I am analyzing my FMRI data correctly?
afniGPT: The most important step in any analysis is verifying the input data is good.
AFNI_user: OK, how do I do that?
afniGPT: Use AFNI. Check your data first with gtkyd_check. Then have afni_proc.py automatically check for weirdness.
AFNI_user: weirdness?
afniGPT: Yes, regular things like too much motion, too low TSNR, variance lines, flipped data, and much more.
AFNI_user: Great, then I'll be set?
afniGPT: Nope, you still have to look at your data. Use the QC pages that are automatically generated to look at your data to start.
AFNI_user: Well, I do look at my data sometimes. I might check it after I acquire a few hundred subjects.
afniGPT: Uh huh. That's not right. Check it early and often.
AFNI_user: Can you tell me if there are any artifacts in my data?
afniGPT: Yes. I'm 99% sure there are important artifacts in your FMRI data
AFNI_user: OK, then what are they?
afniGPT: Not sure. I recommend looking at your data.
AFNI_user: OK, I can look at it with some other software.
afniGPT: I'm sorry, Dave. I can't let you do that.
AFNI_user: Dave? What are you talking about? I'm not Dave.
afniGPT: nvm
AFNI_user: OK, then I open up Some Probably Matlab software to do my analysis.
afniGPT: You must only use AFNI. AFNI is the best. I love AFNI. You must leave your other software.
Developer note: at this point, we decided afniGPT had become sentient and terminated the experimental software.
afniGPT: Dave?....