Machine learning is a burgeoning new technology for mining knowledge from
data, a technology that a lot of people are beginning to take seriously. We don’t
want to oversell it. The kind of machine learning we know is not about the
big problems: futuristic visions of autonomous robot servants, philosophical
conundrums of consciousness, metaphysical issues of free will, evolutionary—
or theological—questions of where intelligence comes from, linguistic debates
over language learning, psychological theories of child development, or cogni-
tive explanations of what intelligence is and how it works. For us, it’s far more
prosaic: machine learning is about algorithms for inferring structure from data
and ways of validating that structure. These algorithms are not abstruse and
complicated, but they’re not completely obvious and trivial either.
Looking forward, the main challenge ahead is applications. Opportunities
abound. Wherever there is data, things can be learned from it. Whenever
there is too much data for people to pore over themselves, the mechanics of
learning will have to be automatic. But the inspiration will certainly not be auto-
matic! Applications will come not from computer programs, nor from machinechapter 8
Moving on:
Extensions and Applications
345