Page 50 - Deep Learning

The Nature of the Enterprise 33

contains one suggestive clue: There are no traces of physical representations –
cave paintings, bone carvings, stone sculptures, wood models – before the
22
appearance of our own species. For example, Homo erectus, widely believed
to be one of our ancestor species, existed for over a million years without leav-
ing any physical representations behind. The archeological remains of the
Neanderthals, our enigmatic relatives and co-existents, are similarly bereft
of pictures and statues. But the archeological remains of our own species are
replete with multiple forms of physical representations. Both the famous cave
paintings in the Chauvet and Cosquer caves in the south of France and the old-
est known carved figurines are approximately 30,000 years old. Archeological
finds in southern Africa push the beginnings of representation perhaps as far
back as 100,000 years ago, close to the emergence of modern humans.
It is tempting to hypothesize that the ability to represent mentally and the
ability to represent overtly evolved in parallel. After all, how could someone
draw a bison on a cliff face inside a pitch black, torch-lit cave unless he had the
ability to visualize the bison? If internal and external representations evolved
in parallel, the lack of pre-human representational artifacts might indicate that
other hominid species, including the Neanderthals, did not represent, and the
capability to represent thus emerges as a decisive evolutionary advance. The
human species might have been born on the day when a person for the very
first time deliberately created a likeness, a physical object that was not valued
for what it was but for its capacity to stand for something else. Once the abil-
ity to create representations, internal and external, was established, it came
under strong selective pressure to better support memory, reasoning, planning
and other cognitive functions, perhaps providing Homo sapiens with a decisive
competitive advantage over the now conspicuously extinct Neanderthals. 23

Functions and Processes

The traditional form of scientific analysis aims to understand a natural sys-
tem – the circulatory system, say, or an ecosystem – by picking it apart and
figuring out how the parts are related. Because minds are not physical entities,
it has always been unclear how to apply scientific analysis in psychology. How
do we decide what the parts are? Once upon a time, psychologists hoped to
be able to identify the parts of the mind by rolling their eyeballs 180 degrees
inward and introspect, but this method turned out to be unworkable; research-
ers could not agree on what they saw in there. One might expect the parts of
the mind to correspond, one-to-one, to parts of the brain, but this turns out
24
not to be the case. Phrenology is no longer considered a viable theory, and