Page 96 - Deep Learning

The Production of Novelty 79

novel products or solutions can emerge only because someone searches longer or
deeper than anybody else. This is indeed how some famous chess-playing com-
47
puter programs win over their human opponents. They search the space of pos-
sible moves and countermoves to a greater depth than human beings can do. in
this view, there is no distinction between analytical and creative processes. For
example, Simon and Craig a. Kaplan claimed that creative problem solving is
heuristic search as usual; a creative solution is distinctive merely in that it depends
on the use of particular heuristics. For example, noticing invariants might be
48
particularly important in solving a puzzle known as the Mutilated Checker Board
49
Problem. This is clearly too specific a hypothesis to apply to a wide range of cre-
ative projects – which invariants had to be noticed to invent radar? – and hence
fails to provide a principled explanation of what is creative about creativity. 50
The notion of heuristics appears to provide a strong answer to the ques-
tion of direction. empirical studies of creative problem solving both inside
and outside the laboratory have verified that the problem-solving heuristics
used by both novice and expert problem solvers can be specified with some
precision. However, because heuristics are task specific, they must have been
51
learned through prior experience. Their application in guiding search in an
unfamiliar problem space is thus a case of projecting past experience onto a
new situation. But the central task for a theory of creativity is to explain how
creative thinkers go beyond experience.
The principle of accumulation also fails to explain why the production of
novelty is limited and difficult. if search were the whole story, there would be
no reason to ever experience an impasse: search, back up, search some more,
and so on, until the solution is complete. indeed, if all it takes to be creative is
to plug along until enough advances have accumulated, then why can we not be
consistently creative? Why do so many garage tinkerers fail to invent anything
useful? Why do so many scientists retire without ever having published a con-
sequential paper, and why do so many painters who paint all their lives produce
nothing but mediocre works? if all it takes to produce a masterpiece is to keep
working, then the production of masterpieces would be routine, unremarkable
and a function of nothing but amount of effort, contrary to all experience.
in short, the accumulation idea explains why novelty is possible but pro-
vides no insight into what is creative about creativity, the direction of creative
projects or the limits on creativity. So-called variation-selection theories of cre-
ativity make no substantive claims but merely rehearse the logical point that
failure must be followed by a renewed attempt. Heuristic search is a powerful
theory of analytical thinking but cannot explain creativity. The central weak-
ness is that Newell and Simon never proposed a theory of the origin of problem