Galatea 2.2, page 13
With my last chapter, the charm broke. I knew it in advance. I saved every trick I had for the end, to break her heart and win her for the present, forever. But of course, a return to feeling only made things worse. I read her the ending. Lovemaking stayed silent this time, skin more a checkpoint between us than a visa in.
“What will you do with it now?” C. asked.
“I don’t know.” The plot had gotten away from me. Escaped its frame. “Send it out, I guess.”
Too rapidly, she agreed. “Of course. You have to.”
Only the usual literary biography would have saved us. Fifteen years of waiting to be taken. Growing stronger, closer to each other on the mound of rejection slips, which we’d have burned for fuel.
The day we heard the book had been bought, we celebrated. Cheer felt forced and punch-drunk. C. assumed the virtue of excitement as bravely as she had managed each chapter up until then. But she was like a mother losing her preschooler to the talk-show circuit.
She tried to show enthusiasm for the production process. She pitched in, but her heart had bolted. She hated those grubbers in New York touching the manuscript, even to typeset it. It killed her to watch those farmers make their way into the brutal market. To see them join the ranks of the century’s displaced.
She would never again listen to a word I wrote without suspicion. Endings, from now on, betrayed her. Simple associative fact: it wasn’t even a question of remembering. What chance does story have against neurons that generalize from a single instance?
The week we learned the book’s publication date, C. received an offer of promotion. The brokerage wanted her to run their Operations cage. The jump, steep and quick in coming, surprised no one but C. She was alone in never knowing how competent she was.
The offer could not have come at a better time. C. needed something, and nothing that I could give. A hurried, three-week trip to Limburg to check on her folks left her edgier than ever. Not even walks worked any longer. A real career might be no more than a changeling baby. But even a changeling can take up the slack of care.
She had some days to decide. We spent them spinning skeins of reassurance. “You’ll be great at it. They wouldn’t have asked you otherwise.”
The day she went in to accept, I prepared a feast. I made decorations. Funny little signs with cartoons of C. on a pyramid of brokers, cracking a whip. Hand-lettered posters reading “Book That Cruise” and “Retirement by 35.”
I could tell, watching her come up the courtyard, that celebration was a horrible mistake. She pounded up the stairs, slammed the door, and held it shut behind her with all her hundred and five pounds, sobbing.
“Beauie, we need to get out of this place.”
I tried to hold her, proffer all the worthless comforts. “Okay,” I managed. “I’m game. Where to?”
The last place. Worse than I expected. “I want to go back to U.”
IMP B ALREADY pushed the envelope. B hadn’t a clue what cats were, or sacks, let alone wives. But it seemed to know how to count them, or not count them, as the case demanded.
If A had been an exercise in verbal pattern recognition, B was a foray into computational linguistics. It knew things like over and under, right of or left of, inside or out. Even that far, I doubted whether it comprehended these containers or whether it just manipulated them cleverly enough to pass. Then again, I began to doubt whether I myself could define the difference.
B could handle syntax. It had a rudimentary sense of the parts of speech and how they operated on each other. And it began to cross the threshold into semantic content. Lentz once or twice tacked on a new subnet to handle different routines—a noun-phrase decoder or a short-term recognition scratch area. In fact, I suppose we were up to Imp B.4 or better.
Lentz assured me that B would handle its own knowledge representations. The frames, the inheritance of classification qualities and exceptions, the scripts: all would fall out as a result of the way B stored associated input. But even in its minute domains, B had to deal with numbingly different kinds of knowledges. With nouns alone, what you could do with “pattern” varied without limit from what you could do with “matching” or “machine.”
I’d lost count of the number of neurodes involved. It had grown big, complex beyond belief. A glitch now set us back whole days at a time. The thing was a monster, distributed, unchartable, out of control. And yet Lentz’s brain, or mine, was hundreds of millions of Imp B’s wide. We could push that matter a little longer, if only just.
We were still experimenting with the size of layers. Bigger was not always better, Lentz told me.
“Life, Marcel, in case you on the humanistic side of the tracks haven’t grokked this yet, involves a series of trade-offs.”
“Yes, we’re onto that.”
“Now. The trade-offs in input layer size. Well, the smaller the layer, the better it generalizes. The larger, the more it can learn to fit into an associative grid.”
“The better it is at generalizing, the worse it is at acquiring new associations?”
“The poet’s a blooming genius. Now we know how you earn your grant money.”
“Does it follow that the more facts it has, the harder it is to take in new facts?”
“Thirty-five is about when that starts to happen, Marcel. You begin to think, ‘Well, I more or less understand how things work. Do I really want to disassemble tens of thousands of tangled, semi-accurate beliefs on the off chance that I might be able to bring one small receptor field into better focus?’”
“Tell me about it. I’m there.”
“Don’t worry, little boy. You’ve a few tricks yet to pick through. And a few years yet to pick through them. I mean, you’re lucky I’ve taken you on. Aristotle wouldn’t accept any student still young enough to have a sex urge.”
“Not a problem, at the moment.”
On days when Lentz engaged himself with design philosophy, he grew expansive, almost pleasant. On days when he built things, he was fun. I tried to keep the soldering gun in his hand and ignore the baiting as much as possible.
“Now: what about the size of the hidden layers? Do you want them bigger or smaller than your input layers?”
“I’m sorry. I give up. We’re going to have to turn all the cards over.”
“Come on, think it out. Consider the translation impedance. Another trade-off. The better the resolution, the more susceptible the net becomes to random noise. Think of B as a curve fitter …”
“That’s all our brains are? Curve fitters?”
“It’s a big ‘all,’ friend. The curve we are trying to fit is as long as existence. As many dimensions. The fact that we can get the infinite data stream to cohere into lumps at all has turned men with as much native intelligence as your friend Plover into mystics.”
“Here we go. Time to slander Harold.”
“It’s not slander. The man makes his claims public. ‘Meanings extractable from a given linguistic configuration may be neither convergent, bounded, nor recursively enumerable.’ Or some such rubbish. He seems to think that because ‘context’ is infinitely extensible, there can be no neurological calculus of interpretation.”
“And you, Engineer?” I tweaked, waiting until he had his hands full of printed circuit and his head deep in the cage. “Do you think that because you are virtuous, there will be no more cakes and ale?”
“Now, Marcel. What in the hell is that supposed to mean?”
“Dunno. Let’s ask Harold.”
“Hn. I’d rather get back to the subject. Sometimes I think the human brain is just one long open parenthesis. So. Tell me. How big do you want your various output layers to be?”
“I guess that would depend on how big an answer we are expecting from any given subnet.”
“Well done. You’re getting cagey. We’ll have you writing NSF proposals in no time.” Lentz’s sarcasms were mellowing with age. “Would you concede, then, that many of our output layers could consist, in theory, of a single neurode, since the cyborgs think that every quest can be rephrased as a series of yes-or-no questions?”
“I wish the lit critters would catch on to that.”
“Yes. Handy, isn’t it? Cleans up a lot.”
“Engineer, can I ask you something? If you’re not a mystic and you’re not a cyborg, what in creation are you?”
“‘Creation’ is a loaded word, Marcel. I guess I’m a lot of little delta rules running recurrently, evaluating and updating themselves.”
“Tell me a different story. I’m not sure I like that one.”
We were well into the millions of connections when B seized up for good. We’d made so many tortuous increments, we’d stopgapped so many glitches that I did not, at first, see this collapse as fatal.
“John is a brother of Jim’s,” I told it. B turned the fact into a stream of hieroglyphic vectors that changed its layout imperceptibly. “Who is Jim’s brother?”
“John,” Imp B replied. Reliant knight. Already it outperformed some aphasics.
“Who is Jim?”
“John’s sister.” That much was fine. I could live with that answer. In fact, it taught me a thing or two about my own presumptive matrix.
I continued, “John gives Jim apples. Who gets apples?”
“Jim gets apples.”
“Jim is given the apples by whom?”
“Jim is given the apples by John.”
“Jim eats an apple. The apple is sour. Jim throws the other apples away. Why does Jim throw the other apples away?”
At that point, B’s cranking time became unendurable. It returned something like, “Jim throws the other apples away because the apples are given by John.”
“No,” I told it, or words to that effect. “Start again. Why?”
“Jim throws the apples away. She does not want them.”
A marginally acceptable answer. Maybe insight hid away somewhere in that tacit implication. But maybe the damn thing was bluffing. Its vagueness depressed me: the slow tyro during story hour, doomed from birth to a career in food service.
“Why doesn’t she want them?”
“She doesn’t eat them. So she can’t want them.”
This alien proto-intelligence differed just enough from sense to make my head throb. Still, we lay within acceptable performance margins. I went on to torture it with, “Jim hits John. Why does Jim hit John?” B had one of its damning seizures. It cranked all afternoon, resetting itself, grabbing randomly at a thousand possible but skewed associations.
Thrashing, it tried a proverb we’d hammered into it the week before. “Jim hit John because one bad apple doesn’t spoil the whole barrel.”
When pushed, it finally failed to answer all together.
Some crucial frameshift fell outside B’s ability to effect. It could not say, John’s apples angered Jim, or No reason, or There could be bad blood I haven’t heard about.
It could not even say I don’t know.
It lacked some meta-ability to step back and take stock of the semantic exchange. It could not make even the simplest jump above the plane of discourse and appraise itself from the air. Although it talked, in a manner of speaking, speech eluded B.
Its brain faltered at that Piagetian stage where the toy disappeared when placed behind a screen. It could not move ideas around. All it could move around were things. And the things had to be visible at all times.
Something was screwy with the way B passed symbolic tokens among its levels. It might grow knowledge structures forever, as fecund as a field tilled with representational fertilizer. But its knowledge about knowledge would remain forever nil. And no patching or kludging on Lentz’s part could set it right. B’s deficiency seemed to be a by-product of the way its constituent nets spoke to one another. The way we’d linked them into the grand schematic.
We postponed the inevitable for as long as we could. I came into the office one evening and found Lentz behind his desk, inert. “I want to change the architecture.” Ahab, well out of port, announcing the slight broadening of plans.
I was too invested to feel demoralized. Maybe I had some naïve image of taking a magnetic snapshot of B and somehow porting it intact into new and more capable digs.
Retrain several million connections from scratch. Had I realized, I might have signed off the project. But I had no incentive to realize. My last month and a half of literary effort had produced no more than half a chapter of train, crossing the snow-lined mountains toward sunny neutrality. I stopped at every other sentence to run to the library and verify that I wasn’t unconsciously plagiarizing. As a result, the thing read like a Samuel Beckett rewrite of the Ancrene Riwle.
U. made changing machine architectures almost trivial. It stretched credibility, but that sleepy hamlet with the two-dollar movie theater and the free corn boil at summer’s end also sheltered a National Supercomputing Site. The town, fallen through the earth’s crust into a dimension where nothing had changed since 1970, consequently had a jump on the next millennium, the advantage of the late starter. Four rival pizza parlors, each named “Papa” somebody, each opening whenever it felt like it. Bars where fraternities split Thursday-night twenty-five-cent plastic cups of beer. And the most advanced, block-long cybernetic wonderland that a paranoid race to preserve faltering world dominance could fund.
I imagined my network’s first freshman comp assignment: “Convince a total stranger that she would not want to grow up in your hometown.” Son of B could whistle its answer to that one in the dark.
The latest nationally funded supercomputer to come of age in U. had itself already lived through half a dozen incarnations. Nor was it really a single machine. It was a collection of 65,536 separate computers, chained like galley slaves into inconceivable, smoothly functioning parallel. Depending on the benchmark, the connection monster could outperform any computing assemblage on earth.
The machine was so powerful that no one could harness it. So notoriously difficult was its programming that major scientists and their graduate-student franchises had already begun to flee U. for sites a tenth as potent but at least manageable.
“We’re moving over to the connection monster, Marcel.”
“Lentz, you’re kidding me. This game can’t interest you that much.”
“What? I’m not afraid of that thing. How hard can it be?”
“Hard. Harder than humiliating your colleagues can be worth.”
“You underestimate my colleagues. Besides, don’t forget the undertaking’s empirical interest.”
“Don’t bullshit me, Engineer. This thing is all smoke and mirrors. And you know it.”
But he knew, too, that he had me hooked. I wanted to see the next implementation up and running on a host tailor-made for neural nets. Lentz wrote a proposal credible enough to pass itself off as real science. By that time, the keepers of the connection monster were so hard pressed to salvage their hardware from neglect that they were taking all comers.
The man was dangerous when he had a plan. He meant Imp C to be profoundly different in nature. He wanted to push the notion of the self-designing system up a level. Reweighting prewired connections would no longer suffice. Imp C would be able to strengthen or weaken the interactions between entire distributed subsystems. It would even grow its own connections from scratch, as needed.
Lentz wanted to get hundreds, perhaps thousands of large, interdependent nets up and running at the same time. He saw them passing endless streams of ideational tokens among themselves. The net of networks would churn at all times, not simply responding passively to new data inputs. When input stopped, it would interrogate itself in ongoing, internal dialogue. Its parts would quiz one another, associate and index themselves, even when alone. Imp C would undertake constant self-examination and reorganization.
Lentz meant to distribute these chattering subsystems not just across the connection monster’s 65,536 processors but across other various and specialized hosts. Each task communicated with the others via high-speed fiber-optic cable. C, if it could be said to live anywhere at all, lived spread all over the digital map.
“Keep out of my hair for a few weeks, Marcel.”
“That should be easy.”
Lentz, preoccupied, ignored my crack. “Just until I dig the foundations.”
The order suited me. I had copy to proofread. I’d also agreed to a number of classroom visits, hoping to justify my freeloading existence.
The class visits were an embarrassment for everyone. Students sat polite but stunned in front of me, their desks circled like Conestogas under attack. The look of shame on each face asked how I could have missed the fact that the age of reading was dead. “How do you work? Where do your ideas come from?” they asked, hoping I would take the hint and go away.
I answered as best I could. But I couldn’t take a step toward the first of these questions without lying. Luckily, lies were all they expected.
When class visits ended, I sat in my office, proofing the book I no longer remembered having written. I went through one methodical ruler-line of text after the other until the Primary Visual Area back in my occipital lobe started to bleed.
I tried to read without comprehension. One catches more errors that way. But sense pressed itself on me, kicking and screaming. The book’s style perched on the brink of nervous disintegration, the most depressing fairy tale I’d ever read. And the only person more depressed than I was my editor, sitting on my unfulfilled contract.
I knew the narration meant to be double-voiced with my protagonist, a resident surgeon living an extended mental breakdown. My depiction of a society in collapse now struck me, if anything, as hopelessly tame compared to those vignettes shoring up the hourly news. But as I read, I kept thinking: Someone has to find this author before he does something desperate.
I spent the first week exterminating typos. The second I spent improvising, in the margins of the last pages, some semblance of redemptive ending. I scratched out the manuscript’s final descent. In its place, I had my collapsing surgeon hold out his hands, catch the woman who had failed to save him. I added a last-minute postscript that returned the tale to the light of its source, the reader, untouched by tragedy. And yet the story still radiated a darkness so wide my pupils could not attenuate.
