Page 90 - Deep Learning

The Production of Novelty 73

the assumption of random generation, one would expect the hit rate to be uni-
form across creative agents within a field of activity.
Campbell’s second argument is that creative thought is serendipitous.
Creative people often search without any specific goal in mind, he claimed, and
there are a good many things to create, so although the probability is low of find-
ing the particular thing that the person intended to find, the probability of hitting
upon something of interest might be much higher. This is yet another quantita-
tive claim unsupported by any calculations. autobiographical accounts in which
artists claim that they occasionally engage their medium without a clear goal or
35
purpose suggest that there is a grain of truth in this. But the stronger theme in
creative work is the deliberate pursuit of goals. Darwin’s decades-long struggle to
understand the organization of the living world was no meandering through the
conceptual landscape on the chance that some interesting ideational specimen
might turn up. The Wright Brothers did not tinker without forethought, and
36
neither Mozart, Beethoven nor ravel stumbled on this or that composition by
combining notes without purpose. 37, 38 to the extent that creative processes are
goal-directed, the generative process cannot be random, because such a process,
by definition, is unbiased and so cannot be guided by a goal.
The application of the variation-selection schema to creativity is rediscov-
ered and reformulated time and time again because it is easily observed and it
is intuitively compelling. But variation-selection, as applied to creativity, is not
an explanatory principle but a logical necessity. if a solution does not work,
the problem solver only has two choices: Generate another solution – that is,
vary the approach – or give up. The variation-selection pattern is a logical con-
sequence of fallibility, and it is necessarily the mode of operation of any agent
who persists in the face of failure. The variation-selection schema, extracted
from its biological content, is neither explanatory nor in need of explanation.
What needs to be explained is how, by which processes, new variations are
generated and selected in the search for a creative outcome or product.

accumulation Through Heuristic Search
a creative person at work on a complex problem does not typically try this or
that approach haphazardly. The successive variations have sensible relations
to each other and constitute a development rather than a random series. as a
biological analogy, ontogeny would be closer to the mark than natural selec-
tion. There is a logic to the unfolding, an internal organization that provides
direction. What is generated in cycle N+1 depends on what was generated and
evaluated in cycles 1 through N.