Gallant, Jack
<p>email is gallant@berkeley.edu
 
Jan 04, 2013 11:12 AM Last name pronounced  GAL ant
(hear Jack pronounce it himself in the UCTVPrime  Vision Reconstruction YouTube)
 
Oct 10, 2011 07:47 PM
see Jack's talk to MBC Monday grad seminar below... given today.   CRON appearance.
 
No videos of Jack speaking online.
but there is an audio
 
introducer  (Michael Krasny) says Jack Gallant as in gallant.
 
MRI is low res... about the size of a pea... perhaps a million neurons in that space... also very slow (takes 4 sec to peak).
focused on early vision area only in the clip research.
ENCODING model is really what they are interested as neurobiologists.
primary visual cortex differs by 200% across people....  so, want to study individual people.
 
decoding thoughts of stroke patient in a coma is a potential use in the future.
 
every person will have their own dictionary (translating from the world to voxel activation);
 
They are only decoding the EARLY STAGES of the visual system (not any other areas).
 
vision is hierarchical:   in future, it would be good to decode early (edges and blobs) and later (meaning of objects)
and have them interact;
 
search thru 18M seconds of YouTube video and find the 100 clips that are most similar and then average the movies;
 
in the magnet they showed hollywood trailers to the subjects;
 
jim haxby developed a method for functionally aligning brains (rather than anatomic alignment);
 
used YouTube cuz it was independent of the hollywood trailers that were used for training the models;
 
question abt Paul Eckman:   truth-telling and lie detection using MRIs....
memories are unreliable;
 
reconstructing cubes of space-time;   each cube is only 1 second long  (but lots of frames in a movie in one single second);
 
is MRI safe:  it is magnetic....  seems to be safe... seems to be innocuous;
 
action potential takes about 1 ms to fire;  cud do a great job of decoding... but u cannot do that...
 
 
 
 
 
 </p>