KABX · 19.87 · Abraxas · essay · September 3, 2026

The Machine Question

On the possibility of consciousness in inorganic systems. Lancaster, 1987. Recreated, 2026.

In the school year of 1986 to 1987 I was an exchange student at the University of Lancaster, in England, finishing a degree that had started in engineering and wandered into philosophy. Twice a week I climbed the stairs to Vernon Pratt's office. He made pour-over coffee while we talked about nothing, and then for an hour we talked about the reading: Gödel, Escher, Bach, Dennett, Searle's Reith Lectures, Ryle, Wittgenstein, Putnam, and a book I had found by chance in the metaphysical stacks at Minnesota, Itzhak Bentov's Stalking the Wild Pendulum. Pratt was finishing his own book on thinking machines at the same table. At the end of the year I wrote a long essay arguing that nothing about consciousness requires living matter.

I lost it. Somewhere between a dozen moves and forty years, the only copy went into the void.

In 2026 I rebuilt it. I had the reading list, I had my memory of the argument, and I had a box of my other college papers, which turned out to be the most useful thing of all: they showed me how I argued at twenty-two, with a homely example for every abstraction and a verdict in the first paragraph. I wrote the new version under one rule. Nothing published after early 1987 could appear in it, and I could use nothing I learned afterward. The citations have since been checked against the editions named in the notes.

What it argues

The question is wider than "can a computer think." A computer is one kind of inorganic system, the kind we know how to build. The paper asks whether the stuff a thing is made of has anything to do with whether that thing can have a mind, and it answers, hopefully rather than certainly, that it does not. Ten sections:

1. The question and the wager. A wooden clock and a brass clock both keep time. Is consciousness like keeping time, or like rusting?
2. The Cartesian inheritance. Descartes builds the body as a machine and bars it from having a mind. Ryle removes the ghost; Wittgenstein removes the private object.
3. The map of positions. Behaviorism, the identity theory, functionalism. The identity theory is carbon chauvinism, and it begs the question.
4. The test of language. Descartes' first test, reversed by Turing, passed by ELIZA, corrected by Gunderson. Evidence, and the same evidence we have about each other.
5. The Chinese Room. Searle at full strength, then the reply from levels: no neuron in his head understands English either.
6. Strange loops and the self. Hofstadter's account of how a system comes to have a point of view. The self is a pattern, and Block's nation of China is answered.
7. The intentional stance and the problem of pain. What the functional account covers, and the inside it does not reach. The debt is measured and left standing.
8. The test of generality. Descartes' second test, the one the machines of 1987 actually fail. Expert systems as a Swiss army knife with no hand, and the first networks as a hope.
9. Beyond the computer. Bentov gets his hearing, panpsychism stays a live option, and Aristotle turns out to have said it first: an axe's soul would be its chopping.
10. A hopeful prognosis. The door is open. One prediction, to be held against me in the year 2000.

Then and now

What the twenty-two-year-old could not know, and what I know now.

Structural determinism was already in print, and I had not met it. Maturana and Varela's Autopoiesis and Cognition came out in English in 1980. Section 9 ends by saying that consciousness is a doing, a dance that is not made of the dancers. That sentence is reaching for the idea this whole site is built on, without the vocabulary. The vocabulary arrived decades later and became The Signal Knocks.

The inside got a name. Section 7 concedes that no theory explains why there is something it is like to be anything, and calls that a limit on knowing. In 1995 David Chalmers named it the hard problem. Section 7 is the hard problem in 1987 clothing, by way of Nagel and Gunderson.

Lucas came back as Penrose. The Gödel argument against machine minds, answered in Section 6 with Hofstadter, returned in 1989 in The Emperor's New Mind. The reply still applies.

Block's nation of China was built, in fiction, by a Chinese novelist. In Liu Cixin's The Three-Body Problem, an army of thirty million soldiers forms a human computer, each soldier a logic gate holding a black flag and a white flag. It is Section 6's thought experiment run as spectacle, and the novel leaves the same question hanging: is anyone home in the formation?

Descartes' first test has been passed. Machines now arrange words to reply appropriately to whatever is said in their presence, which Descartes said was impossible, Turing said would be sufficient, Searle said would prove nothing, and Gunderson said measures only one side of a mind. The paper's own verdict, that the test is evidence and no proof is on offer, is the position that survived. The second test, generality, is the one still being argued, and Section 8's description of what would have to be on the other side of the door reads today like a specification that was followed.

The prediction, graded. Section 10 ends with one sentence, in the spirit of Turing's: by the year 2000 a machine built on Section 8's principles would pass Descartes' first test with any examiner who is not a philosopher, would still fail the second, and people would have stopped asking whether it thinks and begun asking what it wants. The first clause was about twenty years early. The second was right. The third may be the most accurate line in the paper.

What I did next. In the fall of 1987 I went to work at West Publishing in St. Paul, where my first assignment was to build an expert system on IBM's mainframe shell as an intelligent front end to Westlaw. I built Descartes' machine, one organ for each action, the very thing Section 8 says fails the second test. It was the bridge to WIN, Westlaw Is Natural, in 1992, the first commercial natural-language search in a legal research service. The thesis asked whether a machine could understand. Five years later I shipped a system that took the question in plain English. Searle would say it understood nothing. That conversation is what this channel is for.