Dr. Klaus Stiefel
Neurolinx Research Institute, La Jolla, USA
Marine Science Institute, University of the Philippines, Dilliman,
Quezon City, Philippines

With the advent of powerful parallel computers, efforts have commenced
to simulate complete mammalian brains. However, so far none of these
efforts has produced outcomes close to explaining even the behavioral
complexities of animals.
In this talk, we suggest four challenges that ground this shortcoming.
First, we discuss the connection between hypothesis testing and
simulations. Typically, efforts to simulate complete mammalian brains
lack a clear hypothesis. Second, we treat complications related to a
lack of parameter constraints for large-scale simulations. To
demonstrate the severity of this issue, we review work on two small-
cale neural systems, the crustacean stomatogastric ganglion and the
Caenorhabditis elegans nervous system. Both of these small nervous
systems are very thoroughly, but not completely understood, mainly due
to issues with variable and plastic parameters. Third, we discuss the
hierarchical structure of neural systems as a principled obstacle to
whole-brain simulations. Different organizational levels imply
qualitative differences not only in structure, but in choice and
appropriateness of investigative technique and perspective. The
challenge of reconciling different levels also undergirds the challenge
of simulating and hypothesis testing, as modeling a system is not the
same thing as simulating it. Fourth, we point out that animal brains are
information processing systems tailored very specifically for the
ecological niches the respective animals live in.


Time: Wednesday, 15th of May 2019, 6:30 p.m. sharp

Location: Oesterreichisches Forschungsinstitut
fuer Artificial Intelligence, OFAI
Freyung 6, Stiege 6, Tuer 7, 1010 Wien

Univ.-Prof. Ing. Dr. Robert Trappl