“But isn’t that pretty much the way human memory works?” Prudence asked. “And here’s another thing: It produced the right answer at the translator. The right answer.”
Bickel looked at her, turning that fact over in his mind. She was right, by God! And not for the reason he had so glibly spouted.
The thing had produced the right answers in spite of errors and misprogramming. The processing procedure stank. It was heuristic and should not under any circumstances have yielded the desired output.
But it had. Why?
Bickel experienced a mental sensation as though his mind lurched. It was so much like a physical sensation he wondered that the others didn’t notice.
The beautiful clarity with which he understood what had happened in the computer washed through him like a stimulant.
Didn’t the others see it?
He looked at Prudence, at Timberlake, realized this had all occurred in a fraction of a second.
“For motion produceth nothing but motion.”
The words rang through his mind, producing awe at the way apparently disconnected bits—a line of poetry here, a technical phrase there—could link with a simple turn of mathematics to produce a right answer in his mind.
Just the way it had happened in the computer.
Prudence, correctly interpreting Bickel’s expression, spoke quietly, “You’re onto something, John.”
He nodded. “Prudence, you’re our mathematician. What’s pi?”
She stared at him, puzzled.
“I’m serious,” Bickel said.
“The ratio of the circumference of a circle to its diameter,” she said. “A rational approximation would be approximately twenty-two over seven. A closer approximation would be three hundred and fifty-five over a hundred and thirteen.”
“For most applications, that approximation of pi would give us significant results?” Bickel asked.
“You don’t have to ask that. You know it would.”
“Okay, now tell me why you didn’t answer my question by saying pi is a sweet concoction of starchy crust enclosing a filling often of fruit?”
She saw his seriousness in the way he stared at her, waiting. This bore on the problem in some way. She looked at Timberlake and he interpreted her motion as an appeal for help.
“It’s obvious,” Timberlake said. “You set up a category first by saying, ‘You’re our mathematician.’ Then you asked: ‘What’s pi? You didn’t say: ‘What’s a pie?’”
“Yeah,” Bickel said. “You had two screening references through which to filter the question and come up with the right answer. Then, because you sensed this was a rhetorical question in some way, you didn’t try to explain first that there’s no rational number for pi; you just gave me the rational approximations.”
“Well, I knew I didn’t have to explain that to you,” Prudence said.
“That was category ‘common information,’” Bickel said. “All you had to do was produce the significant answer.”
“Holy cow!” Timberlake exploded, seeing where Bickel was leading them.
“Holy Ox, you mean,” Bickel said.
Prudence whirled, pointed wildly toward the computer panel. “But it wasn’t conscious! It couldn’t have been!”
“It wasn’t conscious,” Bickel agreed. “But first crack out of the box, we’ve produced a significant result. And it was no accident. What can we say about the results of this test? First, we can say that the computer had sufficient information to produce an accurate answer despite errors in the system. Second, we can say that we’ve introduced a new kind of sense data into the system previously called a computer. We can go on calling this a computer, but it’s a step up from ‘computer’ now. It has learned how to use a new kind of sense data.”
Prudence started to speak, stopped.
“Screen everything I’ve said here through field theory,” Bickel said. He grinned at them. “Then remember that we matched three energy sources in the Ox. The integrator there set them up to go out identically. The buffer potential of this storage unit here scattered those pulses through the system. They were divided and re-divided … but wherever they matched they reinforced each other.”
“In itself, the original program pulse was a kind of comparator,” Timberlake said. “The computer could compare for accuracy on the basis of signal strength.”
“And the computer already knew how to compare the AAT signals for accuracy by screening them through a code-matching grid,” Bickel said. “Signal strength was merely another kind of grid.”
“If you’re not too busy congratulating yourself,” Prudence said, “consider how some of those rematched signals must’ve grown in strength. The probability is that some elements of the computer have been shocked out of—”
“We’re still running,” Bickel said, but he spoke defensively, realizing that Prudence was right. There were overload fuses throughout the filter to protect components, but stray signals overriding barrier potentials could have played hob with some of the master programs. He looked at the overhead screen which showed Flattery manning the Com-central board.
Flattery appeared relaxed, but watchful, his gaze traversing the big board.
Damn her! he thought.
One instant everything had been rosy, full of elation that the Ox had come a short step up the ladder—not into consciousness … but toward it. And all she could think to do was throw cold water on them.
Bickel met Flattery’s gaze in the screen. “Have you been listening, Raj?”
“I’ve been listening,” Flattery said.
“Have we gone sour, yet?” Bickel asked.
“You really think I’m this hypothetical human fail-safe device?” Flattery asked, holding his tone to a nice balance between mockery and injured innocence.
He almost goes too far, Prudence thought. If he isn’t underestimating Bickel, he’s pressing the limits. One course is as dangerous as the other.
“You’re the logical candidate,” Bickel said, “but I was asking for your comments on progress.”
Flattery suppressed an abrupt feeling of jealousy. Bickel, in spite of the obvious flaw—and that was a gaping thing—balanced so beautifully. Or … he appeared to balance, which was much the same thing as far as this operation was concerned.
“Ahhh, progress,” Flattery said. “If I understand your original test correctly, the pulse-time distances didn’t check out with the space distances. They weren’t proportional.”
“That’s essentially it.” Bickel wondered why Flattery’s tone made him feel so defensive. “They averaged out to almost a zero product.”
“The artificial nerve nets produce something vaguely equivalent to psychological space.” Flattery paused, scanned the Com-central board, returned his attention to the screen and Bickel. “You can say that the test pulses are more or less like sense data feeding into psychological space—a region somewhat equivalent to what Prudence calls imbedding space. I like her cobweb-and-ink analogy. But there’s a big difference between physical space and psychological space.”
He let it hang there a long time, forcing Bickel to admit a dependency upon someone else’s expertise.
“If you’re going to explain it, then get at it,” Bickel said. There was anger in his voice. He didn’t enjoy depending on Flattery.
“Okay.” Flattery kept his tone even and friendly.
“You can time a signal across physical space, repeat it and get matched results. Any difference will have a positive relationship to a change in distance. But psychological space … now, that’s something else again. The time there could depend on mood. What’s mood, John? Is it a comparison between this and previous experiences of a similar kind? Your pulse-time in psychological space will meet many more variables than it does in physical space.”
“Are you saying we haven’t analyzed our results correctly?” Timberlake asked. He glared up at the screen, feeling that he and Prudence and Bickel were arrayed somehow against Flattery.
r /> “You’re trying to arrive at some proportional comparison between the sense world and the physical world,” Flattery said. “But you can’t use the same rules of measurement. Each neuron in your net will introduce an element of random conduction time and you get more randomness from the similar variation in synaptic delay time. The difference between the sense world and the physical world is the difference between time distance and space distance. And the most casual examination of your setup indicates you’ll have random time distances.”
“Zero by probability,” Bickel said. “That won’t wash.”
“That shot-effect test,” Flattery said, his voice bored, “fired off impulses which weren’t time-regulated. You got a variety of delay times there and in your system. That could average out statistically—by probability mechanics.”
“Over the entire net?” Bickel demanded.
“Why not? The bigger the net, the more likely this is to be true. And your net here took in the entire computer.”
“But we got the right answer from the translator,” Bickel said, his voice pouncing. “Try probability on that!”
“I wouldn’t think of it,” Flattery said. “No more than I’d think of coming to definite conclusions on the basis of one test run.”
Bickel glared at him. “Okay, we’ll run it again!”
“No, you won’t,” Flattery said. “Not without figuring how to isolate your Ox from the computer … and before you think of taking any storage units out of the system, ask yourself which one it’ll be. Will it be a unit protecting the life of someone in the hyb tanks? How about a unit controlling the drive?”
“We can’t tell one from the other without a complete block-sort of the entire system,” Bickel protested.
“Exactly. That shouldn’t take more than eight or nine years—with the manpower available to us.”
Flattery’s argument was unassailable, Bickel knew. This didn’t ease the anger that surged through him at the sight of the man’s coldly superior attitude. Still, Bickel felt they had approached some unspoken, elusive, and vital fact that all of them should recognize. They had approached it and wandered away.