Reading What Machines “Think” - Brain Informatics - page 161

Fig. 1. Observing brains and machines
knowledge produces different brain images representing different activation pat-
terns. For example, different conceptual information such as faces, chairs, and
houses produces different activation patterns [6]. These correlations can be used
for better understanding the way the human brain organizes conceptual knowl-
edge. Yet, an extremely important question is: how far we can go with neuroimag-
ing in understanding human mind using these methods? . With the parallelism
between brains and machines we have a wonderful opportunity to answer to this
question. On the brain side (Fig. 1), we have two known variables, i.e., the re-
quired cognitive activity and the observed activation pattern, and one unknown
variable, i.e., the way the brain is performing the cognitive process. In brain
imaging, the aim is to understand and to model the unknown variable. On the
electronic computer side, there are no unknown variables: the three elements are
completely known. This gives a very relevant test-bed. We know exactly how
“ knowledge ” is processed in computers and we know exactly the “ cognitive pro-
cess ” we ask machines to do. If we succeed in studying the correlation between
the cognitive process and the activation image in the electronic computer side,
we can be confident that the same method can be used on the brain side. We
can also answer two additional questions. On the electronic computer side, we
can study if better activation image interpretation models produce better corre-
lations between activation images and cognitive activities. For example, what is
the effect of knowing that processes are stored as code and data? Does it help
in determining the correlation between activation images and the process “ cog-
nitive ” activity? Answering these kinds of questions on the electronic computer
side can help in determining if clearer separations between brain images related
to different cognitive activities correlate with better understanding of the brain
cognitive processes.