[ont.events] AI/Theoretical Aspects/Systems Seminar - Massive Parallelism in Nature and in Computer Sci. by J. Feldman, Dept. of Comp. Sci., U. of Rochester.

voula@utcsri.UUCP (Voula Vanneli) (03/19/85)

                   UNIVERSITY OF TORONTO
               DEPARTMENT OF COMPUTER SCIENCE
          (_G_B = _G_a_l_b_r_a_i_t_h _B_u_i_l_d_i_n_g, _3_5 _S_t. _G_e_o_r_g_e)

ARTIFICIAL INTELLIGENCE/THEORETICAL ASPECTS/SYSTEMS  SEMINAR
             Thursday, March 28, 11 am, GB 220
                  Professor Jerry Feldman
     Dept. of Computer Science, University of Rochester

  "Massive Parallelism in Nature and in Computer Science"

                          Abstract

     Human brains made of millisecond  components  (neurons)
can carry out complex perceptual tasks in less than a second
i.e. in about a hundred sequential time steps.  This  compu-
tational  constraint,  among others, suggests that the algo-
rithms employed by nature are quite different from those  of
conventional  AI.   Several  groups  have been exploring the
direct use of "connectionist" computational models and  have
obtained  some  promising results.  The talk will describe a
model of massively parallel computation, its application  to
problems  of  vision and language, and some of the issues it
raises for theoretical and systems work on parallel computa-
tion.