Carolyn Kellogg, Los Angeles Times, Published June 16 2012
Robot is constructed in sci-fi author’s image
“How to Build an Android: The True Story of Philip K. Dick’s Robotic Resurrection”
By David F. Dufty
Henry Holt; $26
In 2005, at the edge of robotics, a curious project took shape: an android that was an eerily realistic copy of science-fiction author Philip K. Dick. It looked like Philip K. Dick. It wore his clothes. Using sophisticated software that drew upon hundreds of interviews as well as Dick’s many published books, it even talked something like Philip K. Dick.
The android could smile, raise an eyebrow, frown and, most important, interact in conversation. At a few public exhibitions, it stole the show. It won an award at a robotics competition held at Carnegie Mellon University. And then, on an airplane headed to California, the android’s head went missing.
It was, perhaps, the most Philip K. Dick of endings to a very Philip K. Dick story. The chronicle of how this improbable robot was built and came to its unfortunate end is a fascinating one, but David F. Dufty isn’t quite up to the task of telling it in “How to Build an Android.”
Dufty was a postdoctoral fellow at the University of Memphis when the android’s software was being written there, and in reaching across the scientist-lay reader divide he often over-explains the obvious while leaving out the good stuff. For example, the work of android makers who faced a nearly impossible deadline gets an all too brief summary, but Dufty spends three pages explaining to us what Google is. I’m fairly confident that to anyone who is curious about Philip K. Dick and androids, Google needs no introduction.
It was while on its way to a demonstration at Google that the android lost its head. It was in the care of David Hanson, a sculptor turned roboticist and the man who’d had the idea of making a Philip K. Dick android in the first place. Exhausted from robot exhibitions and not realizing what he’d done, Hanson left the head behind during a flight transfer – giving new meaning to the phrase “overhead bin.” It was briefly spotted but never recovered.
Hanson was trained first at the Rhode Island School of Design, and he made his way into robotics through mechanics. He wanted to create human-like faces that could mimic human emotions. Instead of making a robot with roughly human outlines like C-3PO from “Star Wars,” he constructed android heads with plastic skulls and mechanics that could approximate muscles. He even developed his own rubber-like material for the skin, which he continued to refine as the Dick android was built (and for the Albert Einstein android after that – but that’s another story). Hanson had impressed the likes of “Wired’s” Chris Anderson, but he was working largely on his own in Texas while finishing a doctorate in interactive arts and engineering/aesthetic studies.
It was only after being connected with the staff of the Institute for Intelligent Systems at the University of Memphis that the idea of the Philip K. Dick robot came together. For many years the institute had combined computer programming, psychology and linguistics to push forward how computers can interact.
Most of the IIS work had been focused on education; one major project created on-screen teachers that could, with increasing competence, evaluate and respond to real-life students in learning situations. This kind of interactivity now can be seen in the Internet helpers popping up on some corporate commercial websites (although, if it’s derived from the IIS or was developed in parallel elsewhere, this book doesn’t say). Like those Internet helpers, IIS’ teachers responded to typed queries. Hanson’s work sparked an idea: What if they could move all that teacher programming into a physical android?
The intellectually nimble programmer Andrew Olney took charge of the software side. He was a leading doctoral student in the institute labs, and he connected well with Hanson. Hanson wanted to create a specific human prototype, such as Philip K. Dick; Olney liked the idea. However, financial supporters didn’t – so as the two worked, and applications for funding failed, the android lost its didactic mandate. It wouldn’t be a teacher, it would be Philip K. Dick, an iconoclastic, hard-working pulp writer with a brilliant, occasionally paranoid imagination that could veer into the stratosphere.
With the help of one of the people involved with “A Scanner Darkly,” the 2006 film based on Dick’s novel and in production at the time, the android creators met and got the buy-in from Dick’s estate. It was support like this that kept Hanson, Olney and the rest, who were essentially volunteering their time, enthusiastic about the project. Dick’s daughters sent them an outfit that he wore before his death in 1982; his speeches and books were parsed and input. He was loaded with speech-recognition software, and Olney wrote code that allowed the android to learn from conversations, recognizing repeated questions and refining responses.
It’s at this point that the biggest questions come up: How close was the process of the android Philip K. Dick to actual thinking? Can artificial intelligence rival human intelligence? (Some say it already has.) With Hanson’s painstakingly adjusted facial robotics, “Dick” could express what appeared to be emotions, which were keyed to his internal processes. But that’s not actually being emotional – or is it? Unfortunately, these questions are only glancingly addressed in “How to Build an Android.” As I read, I kept thinking of the public radio show “Radiolab’s” outstanding ability to tackle big-picture questions while explicating scientific detail – there isn’t enough of either in this narrative.
The one thing Dufty has going for him is that he was studying at the University of Memphis at the time, so he knows the atmosphere of IIS from the inside. He ably describes the fertile, feverish atmosphere of intellectual endeavor, the kind of place where a crazy idea – like building a Philip K. Dick android – could take hold.