Page 163 - Deep Learning

146 Creativity

The scaling principles at work in replicating the alternations in mode and
tempo at longer durations are quite general: As a chain is no stronger than its
weakest link, so a stack of subgoals is no more realizable than its most intrac-
table element. A process composed of multiple events that are randomly dis-
tributed over time, and each of which exhibits alterations in mode and tempo,
will itself exhibit such alterations.

Other Properties That Scale

If a creative project requires multiple insights, each of which is massively
contingent and hence unpredictable, the progress of the project as a whole
becomes unpredictable. This contributes to the open-ended flavor of creativ-
ity. It also has three specific consequences that are consistent with what is
known about creative work: First, the uneven rate of progress is an incen-
tive to work on multiple projects in parallel. While there is a temporary
lack of progress on one project, it might be possible to make progress else-
where. Creative people engage in a network of semi-independent but related
enterprises. 29
Second, creatives are unable to keep deadlines. If they cannot predict
when impasses will appear or how long it will take to overcome them, they
cannot promise to have a product or solution ready by a given date. Past expe-
rience might have taught them that there is a high likelihood that current and
future impasses will be overcome, but this expectation does not provide a way
to predict when a particular breakthrough will come – hence, the apparent lack
of ability to keep deadlines.
Third, creative individuals are unlikely to flourish in bureaucratic orga-
nizations with top-down goal setting and scheduling. All three features are so
commonly noted that they are part of both folklore and systematic studies of
creativity. The important point for present purposes is that all three features
follow directly from the massive contingency of impasses and the random dis-
tribution of the time to resolution, principled consequences of the cognitive
processing principles laid out in Chapter 4.
If impasses are caused by the activation of knowledge that will constrain the
solution space inappropriately, then the role of expertise in creativity acquires
a double edge. On the one hand, more prior knowledge will help constrain
30
the search space for the analytical part of the work. On the other hand, more
extensive but inappropriately activated prior knowledge might constrain the
search space in a way that interferes with the solution. The implication is that
expertise exhibits a particular relation to success in creative endeavors: At low