On Intelligence is the best-selling book by Jeff Hawkins, the man who created the PalmPilot, Treo smart phone, and other handheld devices. It’s an ambitious book, an attempt to reconcile our knowledge of what the brain is with how we program and talk to computers, and attempts to create a grand unified theory of the mind.
The ambition is on display early, when Hawkins declares “This theory of the brain can help explain […] how we are creative, why we feel conscious, why we exhibit prejudice, how we learn and why ‘old dogs’ have trouble learning ‘new tricks.’” (pg 6)
Like founder and former president of MedZilla Dr. Frank Heasley, Mr. Hawkins got his start at UC Berkeley. At Berkeley, Hawkins had his first breakthrough – that we program machines to do things, and judge them (and people) on what they do. For instance, R2-D2 does astrogation in Star Wars. But it is not what people do that indicates they have a brain, but rather what they think, which Hawkins believes should be the goal of technologists at work today – create computers that think, not do.
The brain is not a computer, but a memory system that stores experiences in a way that reflects the true structure of the world, remembering sequences of events and their nested relationships and making predictions based on those memories. For instance, as on page 56, Hawkins shows how seeing a dog, hearing a dog bark and petting a dog all operate on different areas of our brain, yet all lead back to instructing us to go “oh, this is a dog!” This is an (admittedly simplistic) example of the sort of neural networked pathways he encourages scientists to devote further study to – not the differences between the different areas of the brain, but how they work similarly.
Hawkins leans heavily on describing heuristics without naming them and talking up the brain as a predictive organ, asserting that that is the key. He describes a thought experiment about changing the location of the knob on your home door – how would a computer be able to tell? Historically, we’d maybe have to program everything about the door – it’s dimensions, weight, color, etc., which isn’t feasible. But your brain makes predictions – the handle should be here, the door should close with this much force, and these constant predictions are what make the mind great. Similarly, the brain predicts the existence of a dalmatian based on the data in an image like this:
Chapter Six is the densest chapter, going into the actual biology of the human brain. It’s a little challenging to those not steeped in neuroscience; fortunately, Mr. Hawkins is able to walk the average reader through it through clever graphs that explain the basics clearly.
Hawkins’ conclusion, from his description of how the mind and intelligence works (as a heuristic prediction machine based on previous knowledge) gives a dim view of us getting artificial intelligence on par with HALor Lieutenant Data from Star Trek; while the capacity issue has largely been overcome in the intervening decade since, the connectivity required to link all the “neurons” of an artificial intelligence requires essentially quantum computing on a grand scale, which is still some distance off.
Hawkins leaves us with this hopeful note, though, on page 222:
“I am less interested in the obvious applications of intelligent machines. To me, the true benefit and excitement of new technology is in finding uses for it that were inconceivable before”
The idea that we just don’t know what we don’t know is absolutely thrilling, and gives this book an upbeat coda heading into the 21stcentury.