By Wray Herbert
A NEW BOOK EXAMINES HOW
KNOWLEDGE OF THE BRAIN'S WORKINGS CAN HELP US
BUILD SMART MACHINES.
How a New Understanding of the Brain Will Lead to the
Creation of Truly Intelligent Machines
By Jeff Hawkins and Sandra Blakeslee; Times Books, 2004
readers of this magazine know that AI stands for "artificial
intelligence," and that there is a vast academic
field, decades old, devoted to researching and developing
AI, otherwise known as "smart machines."
Well, introduce yourselves to RI. RI stands for "real
intelligence," and it is the coinage of Jeff Hawkins,
author of a new book, co-authored with New York
Times science writer Sandra Blakeslee, called On
Intelligence: How a New Understanding of the Brain Will
Lead to the Creation of Truly Intelligent Machines.
The title, though cumbersome, is a concise summary
of the theory Hawkins offers up for consideration by
the scientific community. An electrical engineer by
training and a computer architect by reputation, Hawkins
is probably best known for inventing the original Palm
Pilot and Handspring. But what many don't know
is that he also created the Redwood Neuro-science Institute
in order to encourage research on fundamental brain
functions like memory and cognition. Indeed, Hawkins
says that his computer engineering achievements have
really been offsprings of his true passion: understanding
the basics of how brains work, and why the human brain
is more intelligent than those of other species.
On Intelligence offers his grand theory.
In a nutshell, Hawkins argues that AI doesn't
make sense. The whole enterprise is based on the notion
that the way to make a smart machine is to make a fast
machine. But machines are already faster than human
brains and they still can't compute simple everyday
things, like whether to grab your umbrella on your way
out if the weather is threatening. Or, to quote Hawkins
directly: "Why can a six-year-old hop gracefully
from rock to rock in a streambed while the most advanced
robots of our time are lumbering zombies?" For
that kind of "real" intelligence, Hawkins
says, you need something far more sophisticated than
more and more computations of incoming data.
What is needed, Hawkins argues persuasively, is a
machine that does what human brains do so well. They
tap into memory and use stored knowledge to predict
the future. Memory and prediction are the core elements
of RI theory, and the neuroscience underlying this theory
focuses on the neocortex. Every creature, from a lowly
snail to a theoretical physicist, uses recall of past
experience to determine how it will behave in the present.
Yet humans have such an evolved and flexible neocortex
that they can make adjustments in real time, not in
evolutionary time, like the snail. At least that's
the theory. And, Hawkins is quick to concede, it is
a rudimentary theory at this point, still in need of
Indeed, Hawkins's Redwood Neuroscience Institute
is the only neuroscience center dedicated to studying
the neocortex and RI theory. Others will surely challenge
and try to disprove Hawkins's ideas. But after
reading On Intelligence, one gets a clear sense
that that is what Hawkins would welcome.
Wray Herbert is a freelance writer on culture
and society, and is based in Washington, D.C.