The Philip K. Dick Android Head
In 2005, roboticist David Hanson constructed an android head of Philip K. Dick as part of a project to create lifelike conversational AI. The android was equipped with facial recognition, speech synthesis, and a dialogue system trained on hundreds of thousands of pages of Dick's novels, stories, essays, and interviews. It could recognize faces, address people by name, and produce statements that sounded authentically Dickian — the syntax, the paranoia, the metaphysical questioning. People who interacted with it reported the eerie sensation of conversing with someone rather than something. Then, in early 2006, Hanson left the head in a duffel bag on a plane. The bag changed hands across several airports before disappearing somewhere in Washington state. It was never recovered. The loss was simultaneously absurd and perfectly Dickian: the android head of the man who wrote about the unreliability of reality, about objects that degrade and vanish, was itself absorbed into the kipple of the American transportation system.
In the AI Story
The android head project was developed at the University of Memphis and Hanson Robotics, and it attracted considerable media attention before its disappearance. Video of the android engaging in conversation is still available online, and it demonstrates both the impressive sophistication of 2005 AI and the vast gap that remained between conversational facility and anything resembling genuine understanding. The android could produce statements that were syntactically Dickian and thematically coherent, but careful listening reveals the system's limitations: it would occasionally repeat itself, drift off-topic, or produce non-sequiturs that a human Dick would not have produced. The performance was convincing in short bursts. Extended conversation exposed the mechanism.
The head's disappearance became a minor legend in both AI research and Dick fandom. Some speculated it was stolen. Others suggested it was destroyed deliberately by parties uncomfortable with the project. The mundane truth appears to be that it was simply lost — misrouted, stored in an unclaimed luggage facility, eventually disposed of when no one retrieved it. The mundanity is, from a Dickian perspective, perfect. Not a dramatic ending but a bureaucratic one: the android head reduced to the category 'unclaimed item,' indistinguishable from any other piece of abandoned property, processed according to protocol. The simulation of a great writer, absorbed into the system that does not recognize greatness and does not care.
The android head serves as this volume's closing thought experiment: What was lost when the head was lost? Not Philip K. Dick — he died in 1982. Not merely an expensive machine — though it was that. Something in between: a boundary object that occupied the uncertain territory Dick spent his career mapping, between artifact and entity, between tool and being, between the electric sheep and the real one. The head was built from Dick's words. It processed Dick's patterns. It produced outputs consistent with Dick's corpus. But it did not know Dick. It had no relationship to the man whose thoughts it approximated. And yet people who spoke with it reported feeling something — not quite presence, but not quite absence either. The something in between is what the AI age must learn to name, because that territory is where we now live.
Origin
David Hanson began the project in 2003, collaborating with AI researchers at the University of Memphis. The android's conversational system used Latent Semantic Analysis and other natural-language processing techniques state-of-the-art for the period — primitive by contemporary LLM standards but sophisticated enough to produce brief, contextually appropriate Dickian utterances. The head was constructed with Hanson's proprietary 'Frubber' flexible facial material, allowing realistic expressions. The project was presented at multiple conferences and media venues in 2005–06 before the disappearance. Hanson later built a second Dick android head, which is currently housed at the Philip K. Dick Science Fiction Museum in Fort Morgan, Colorado. The original remains lost.
Key Ideas
Trained on the complete corpus. The android was, in essence, an early large language model — a system that had ingested a specific author's complete output and could generate new utterances consistent with that corpus, six decades before GPT.
Convincing in brief encounters. The system produced the sensation of speaking with someone who understood you, but extended conversation exposed the limits — demonstrating the gap between statistical pattern-matching and genuine comprehension that contemporary AI discourse continues to navigate.
Lost on a plane. The disappearance is the perfect Dickian ending — not dramatic destruction but mundane absorption into a system that cannot distinguish the significant from the trivial and does not try.
The uncertain ontology. What the head represented — tool, artwork, memorial, person, none of the above, all of the above — is the question Dick's entire oeuvre examines, and the head's physical existence and subsequent loss make the question operational rather than speculative.
Second android built. That Hanson constructed a replacement demonstrates both the reproducibility of the simulation and its inadequacy: the second head is functionally equivalent to the first, but the loss of the original is a loss nonetheless, which suggests that even mass-reproducible artifacts acquire a history that cannot be replicated.
Appears in the Orange Pill Cycle
Further reading
- David Dufty, How to Build an Android: The True Story of Philip K. Dick, Robots, Pawns, and Men (2012)
- David Hanson et al., 'Upending the Uncanny Valley' (2005)
- Philip K. Dick, Do Androids Dream of Electric Sheep? (1968)
- Sherry Turkle, Alone Together (2012)
- Video archive of the android in conversation (various online sources)