Page 190 - Deep Learning

The Growth of Competence 173

mock performances as a preparation for the real thing, must have had a strong
impact on the survival rate of a species that relied on acquired rather than
innate competence. But practice requires no genetic mutations, no changes in
brain or body – the person who practices a task performs the same actions as
the person who performs that task for the sake of its results – merely a change
in how the activity is labeled by the surrounding community. The practice of
practicing is perhaps humankind’s greatest invention.
9
practice is extraordinarily effective. Repeated attempts to perform a
task automatically produce improvement. Consider the following thought
experiment: imagine a pretext for asking some person to fill out an unfamil-
iar, made-up but official-looking form that requires the insertion of various
pieces of information here and there, simple arithmetic operations on certain
numbers, the results to be inserted in the appropriate boxes (inevitably located
on a different page), conditional rules as to which part of the form to fill in
next – in short, an unfamiliar version of the kind of bureaucratic product that
is all too familiar. After laboriously filling out the form, our unwitting experi-
mental subject is asked to wait a few minutes. He is then informed that the
experimenter’s dog ate the form; would he kindly fill it out again? We repeat
this charade 10 times. in this situation, the person has no reason to learn, but
we need conduct no study to know what will happen: The time it takes him to
complete the form the 10th time will be a fraction of the time it took him the
first time, his behavior will be less hesitant and he will make fewer mistakes
in following the instructions. That is, he will be on his way to mastering the
task of filling out our nonsense form. The repeated execution of a coordinated
series of actions is sufficient to trigger change, at least in the early stages of
learning. improvement is an almost automatic side effect of activity.
in terms of the magnitude of the improvement, practice is the most powerful
manipulation of behavior known to cognitive psychologists. in the long run, it
10
brings about changes in behavior that are much greater than the typical effects
caused by other types of manipulations such as alterations in the task itself or
in the task instructions, if any. Extended practice can generate a level of perfor-
mance that is not merely superior to initial performance, but orders of magnitude
superior. Some chess masters can play blindfolded against as many as 30 average
players simultaneously and win almost every match. The precision of a world-
class v iolinist or bass player in hitting exactly the right note at the right time is a
source of admiration by even professional musicians. performance improvements
of several orders of magnitude are routinely observed in training experiments.
The familiarity of practice makes its causal efficacy seem self-explanatory:
A learner practices in order to improve, and he improves because he practices.