Undeniable: my ability to recall was not what I remembered its having been. Soon I would forget even the lovely heft of forgetting. The command to picture was, like that departing train itself, heading south. The line sounded its last call over the PA, a Transalpine Intercity just pulling out of the station.
My brother Russie called, from Florida. He almost never did. He checked to see how I was surviving bachelorhood. I asked him if he remembered a book. Something our mother might have read to us once.
"Mom? Our mom? We talking about the same person?"
"Come on. She read to us all the time. She taught me how to read."
"Don't know about your mother, but mine used to surface-mount me to the cathode-ray tube."
"She did not. We were strictly rationed." The last children in America who had to ask to turn on the set.
"Maybe that was the case for you guys."
"Well, you were younger than we were," I conceded.
I could hear Russ's askance silence on the other end. Do we need to call the pros? Start the intervention? "Yeah, bro. I was younger than you."
I couldn't imagine where else the line might have come from. Imagine a route, stretching out leisurely to the south. The day is bracing. It should not be this crisp yet, this chilly or sere. The train eases to life. It builds steam. It stokes up, unfolding itself along that first great curve, and leaving becomes real.
No question: it left something chest-tightening behind, tracking out to the taunting horizon. The book I wanted to write, the book I must have heard somewhere in infancy, unrolled farther than I could see from the locomotive, even leaning out dangerously to look. That passage stretched out longer than love, longer than evasion, longer than membership in this life. It lingered like a first lesson. Outlasted even my need to pin down the broken memory and reveal it.
That sentence, the return leg of the northern line, tried to leap the tracks of desire. But it needed me in the tender. I was the free rider, allowed to hitch on to my phrase's urgent run provided I kept the throttle open and the boiler stoked. But the low-grade hammering on rail that haunted my livelong day dampened to a distant click the longer I failed it. I felt myself dropped off at a deserted siding, numb and clueless as to how I had arrived there.
I sent Lentz a return Post-it. The things were pernicious: just enough adhesive for a temporary stick. "Thanks for the present. No apologies necessary. But please: no more literature. What I need is offprints."
I had but to say the word, and offprints poured in. Lentz sent me articles he had written. He sent me pieces compiled by colleagues and competitors. He forwarded stuff almost a decade old, and submissions that had not yet seen the light of publication. He attached no more notes. If I wanted to read his work, he would not object. But he wasn't going to tutor me.
Neural nets, I learned, had a way of casting themselves over people. According to the miraculous drafts Lentz sent me, there was no hotter topic. Researchers across the whole spectrum of disciplines emptied deep pockets into the promising tangles of simulated brain.
In a previous life, I had brushed up against machine intelligence. For a few months, I wrote code that turned consumer goods artificially lucid. I worked for an outfit that wanted to make household devices savvy enough to anticipate needs that potential purchasers didn't even know they had.
I made appliances expert in their own use. I built the rule base and tuned the reasoning. I linked a table of possible machine states to a list of syllogisms that told the device how to respond in each case. I hooked the device to sensors that bathed in a stream of real-world data. These dispatches threaded a given appliance's inference engine like rats in a behaviorist's maze.
If the data found their way to an exit, they became conclusions. All that remained was the gentle art of interface. I got the device to weigh the situation. I instructed it to say, "Hey! You sure you want to do that?" or "Let's try that again on 'Puree,' shall we?" Convincing the user was the delicate part, far harder than getting the device to reach its decision.
My expert systems couldn't be called intelligent. But they did get me thinking about what could be. I thought about the question for a long time, even after I jettisoned the commercial interests. What was memory? Where, if anywhere, did it reside? How did an idea look? Why was comprehension bred, or aesthetic taste, or temperament?
Predicates threaded my neural maze. After great inference, I came to the conclusion that I hadn't the foggiest idea what cognition was. Nobody did, and there seemed little prospect of that changing soon.
No tougher question existed. No other, either. If we knew the world only through synapses, how could we know the synapse? A brain tangled enough to tackle itself must be too tangled to tackle. Tough, too, to study the workings of a thing that you couldn't get at without breaking. I guess I gave up thinking about thought some time around my thirtieth birthday.
Something about the basic debate upset me. On the one hand, philosophers maintained that the only way into the conceptual prison was introspection. This drove empiricists up the cell wall. Tired of airy nothings, they spent their time amassing chaotic libraries of unrelated data down at neurochemical level.
The top-down thinkers fought back: because thought played a role in experimental design and interpretation, neuroscientists undercut their own efforts. Cognition compromised itself. Recursive by nature, mentation wasn't going to yield to measurement alone.
Cognitive science seemed to me deadlocked. But overnight, while I was away, everything changed. The impasse broke from both ends. Smart appliances kicked out the jambs. The low-level wetware workers came into instruments that allowed them to image the omelet without breaking an egg. At the same time, the top-down people hit upon their own leverage, the neural nets that Lentz's snarl of articles described. Connectionism.
The young connectionist Turks lived on a middle level, somewhere between the artificial-intelligence coders, who pursued mind's formal algorithms, and the snail-conditioners, who sought the structure and function of brain tissue itself. The Center's warrens sheltered all species. But in the halfway world of neural nets, the point man at this place seemed to be Dr. Philip Lentz.
The new field's heat generated its inevitable controversy. I sensed a defensive tone to many of Lentz's publications. Both the neural physiologists and the algorithmic formalists scoffed at connectionism. Granted, neural networks performed slick behaviors. But these were tricks, the opposition said. Novelties. Fancy pattern recognition. Simulacra without any legitimate, neurological analog. Whatever nets produced, it wasn't thought. Not even close, talk not of the cigar.
In his articles, Lentz took these accusations and ran with them. The brain was not a sequential, state-function processor, as the AI people had it. At the same time, it emerged to exceed the chemical sum passing through its neuronal vesicles. The brain was a model-maker, continuously rewritten by the thing it tried to model. Why not model this, and see what insights one might hook in to?
Having stumbled across connectionism, I now couldn't escape the word. I heard it in the corridors. I nursed it at Center seminars, seated in the back for a quick exit. I read about it throughout the worldwide electronic notefiles and in the stack of diversionary texts that replaced my nightly dose of forgotten fiction. Neural simulation's scent of the unprecedented diffused everywhere. I followed along, moving my lips like a child, while Lentz declared in print that we had shot the first rapids of inanimate thought.