Page 208 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

198 CHAPTER 9 Theory of the Brain and Mind: Visions and History

earlier neural network models. Rather CCN is simply the biological side of
what those of us who were involved in the founding of INNS and related institution
envisioned decades ago. It is the type of modeling we expected as more neuroscience
data became available due to technical advances such as fMRI, and as the models
themselves evolved.

5. DISCUSSION

My bias is toward the type of models based both on neuroscience and on principles
that embody cognitive requirements. These include my own work (e.g., [79e83]).
Yet all of the researchers discussed in the last section are dedicated scientists who
have made serious efforts to model important cognitive and behavioral phenomena
while incorporating sophisticated neural data. Sometimes models generated from
two or more different intellectual sources converge enough to differ mainly in details
about the network roles of specific brain regions, an example being the discussion by
O’Reilly, Frank, Hazy, and Watts [63] of the similarities and differences between
their model of the basal ganglia in conditioning and a model of the same process
by Brown, Bullock, and Grossberg [7].
Hence it is in the interests of the field that different modeling groups interact more
than they have. In particular there need to be more conferences which encompass all
these different modeling communities and thereby facilitate dialogue about the
comparative merits of different models.
The other development that would advance the field is for the scientific community
to accept the existence of a “theoretical cognitive neuroscience” or “theoretical neuro-
psychology” that has a life of its own, interacting with but partially separate from
experimental cognitive neuroscience. There should be more centers devoted to the
theoretical understanding of mind and brain. The Center for Cognitive and Neural
Systems at Boston University is such a place, being over 30 years old and spanning
the biological and engineering components of neural network theory, but I am not
aware of another center like it anywhere in the world. This development would be
analogous to the separate, independent, and interacting existence of theoretical and
experimental physics. Theoretical physics has been a respected subfield at least since
the late nineteenth century with the status achieved in Germany by Hermann von
Helmholtz, Gustav Kirchhoff, and Max Planck [84].
So some changes still need to be made in the cultures of various scientific
subcommunities for INNS’s primary vision to be fully realized. Changes in the
reward structure of science, particularly in the United States, would facilitate this
development. The all-or-none, almost casino-like, nature of grant funding in my
country discourages dialogue between advocates of competing models and drains
resources that might be available for more of the right kind of conferences. Yet a great
deal of good work linking the biological and technological sides of the field has
already taken place, along with an explosion of relevant journals and both large
and small conferences. So the future of the field has a good chance to be bright.