FW: BRAIN MODELS

Dear Jeffrey,

A few thoughts as a follow-up to our weekend discussion.

We need to decide between two objectives for the project. The choice of objective will also help determine the criterion(a)  of success for your prize award. These are the two possible objectives:

1) The most successful pattern-recognition model.

2) The most biologically accurate pattern-recognition model.

The two are correlated but not equivalent. The way the brain solves certain problems may fall short of the best engineering solution (e.g. the neuroanatomical blueprint of the brainstem arousal systems would be flunked at any school of engineering; it could be designed with more redundancy to minimize the risk of coma in neurological illness); and the best engineering solutions for certain problems may have nothing to do with the way the brain approaches them (e.g. certain computations are conducted much faster by artificial devices than by the brain).

For objective (1) the criterion could be the broadest range of patterns (both for object recognition and for decision making/action selection) that the system can successfully form and utilize. For objective (2) the criterion could be the best fit with experimental data on pattern formation and use (also both for recognition and for action selection). This will require an experimental arm, most likely with functional neuroimaging.

Yann LeCun's background and current work at Courant is clearly along the lines of (1); it is essentially applied engineering work aiming at "intelligent" robot design. By contrast, Misha Tsodyks' work at the Weizmann Institute is of a very basic computational neuroscience kind aiming at understanding the biological brain, along the lines of (2).

I wonder whether instead of competing groups, one can emphasize competing (or perhaps complementary) approaches, aiming to see which would produce more interesting breakthroughs in some sense. This could entail having two "inhouse" teams interacting both competitively and cooperatively, generating complementary insights - one team working toward objective (1), the other working toward objective (2). I recently heard the term "coopetition" somewhere. The question then arises about my role. I would like to work with both teams. This may sound a bit oxymoronic, but perhaps there is some merit to it. Should you decide to go ahead with the prize idea, perhaps both kinds of teams from other institutions could be encouraged to participate.

I will continue to think about these matters and I assume so will you, Hopefully we will arrive at some clarity soon. Meanwhile, I will be surrounded by people of relevant backgrounds and talents, many of whom I had not met before, at the Kavli Insitute in Santa Barbara much of next week,. These interactions may prompt further thoughts, and maybe even revisions of my proposed team composition.

Best wishes,

Nick

Elkhonon Goldberg, Ph.D., ABPP

Clinical Professor of Neurology
New York University School of Medicine

Diplomate
American Board of Professional Psychology
in Clinical Neuropsychology

315 West 57th Street
Suite 401
New York, NY 10019
USA