Cyber Rogues (77 page)

“A lot of things have been going on that we don’t know about,” Shipley said. “Things that go back to before Tyron even joined the company.”

“Oh? Like what?”

“I’d bet that Tyron had a lot to do with that information at SDC not being made public. He had some kind of deal worked out before he left the SDC—that he’d bring it with him into an area where it can be exploited commercially. Some people are going to make a lot of money out of this, Joe. But it won’t be us.”

It took a moment for Corrigan to see fully what Shipley was saying. “Surely not,” he protested.

Shipley shrugged. “Why do you say that? It wouldn’t exactly be the first time something like that has happened. As a matter of fact, I did some quiet checking on the side while you and Evelyn were away. There are no licenses payable for using the VIV technology that was pioneered at SDC, and I’m pretty sure the same is true for DIVAC. That means that the information can be used freely by anyone now, without restrictions. So Tyron can bring his know-how into CLC and earn himself a lot of gratitude. That’s what it’s all about.”

Evelyn sat back in her chair. “What can you do?” she said. “I guess we’re just a different kind of people. That’s the way things have always been with half the world. Probably they always will.”

Corrigan snorted. “Are you saying we should lie back and enjoy it? Well, you can if you want. But I’ll be hanged if I will.”

“What do you propose?” Shipley asked, not bothering to disguise his skepticism.

Corrigan turned away and banged the side of a steel electronics cubicle with the flat of his hand. “Right now, Eric, I don’t know,” he muttered. “But dammit, I’ll think of something.”

The den of Evelyn’s apartment at Aspinall was darkened, lit only by the green-shaded lamp on the desk. Corrigan stood by the window, brooding to himself as he stared out at the lights of the city. So what, exactly, was Tyron proposing to deliver that was generating so much excitement and attention? he asked himself. Functionally it would still be EVIE. For anyone using the system, the fact that a different behind-the-scenes technology was supporting the vision and acoustics would make only a marginal difference. It was still what Victor Borth had called a “toy”: something that played at imitating the world. That would be of interest to some enterprises, and no doubt Tyron and whoever he was in league with had identified some potential—possibly some quite substantial potential. But Corrigan knew that what the people with the
real
money wanted was something else. Okay, he thought to himself, so those were the rules, were they?

“Joe, are you coming to bed?” Evelyn’s voice said from the doorway behind him.

“Not really sleepy.”

“It wasn’t really sleep that I was thinking about.”

He turned and smiled tiredly in the light at the window. “I have to say my prayers first. You know how the Irish are.”

She came in and moved close to him. He slipped an arm around her. “Still letting it eat away at you?” she said.

“Oh . . . just thinking.”

“You can’t change anything. Start thinking about moving to another job if it’ll help. We’ll manage.”

“Just walk away? Wouldn’t some people like that!”

“I know that the Irish are fighters, too. But you can’t fight this.”

“Well, maybe you’re being just a little bit too quick on handing down that verdict.”

She turned her head and looked at him uncertainly. “Why? What have you got in mind?”

He thought for a second, then said, “Let me check on a few things first, before I start going into it. Okay?”

“If you say so.”

He squeezed her waist and patted her behind through her robe. “Go and get warm, then. I’ll be through in a minute.”

“Hurry up,” she whispered, kissing him on the cheek, and left the room.

It would still be before eleven in California. Corrigan went over to sit down at the desk and called Hans Groener’s personal record onto the terminal’s screen. He selected the phone number and pressed a key to initiate auto-call. Moments later, Hans’s features greeted him. They talked for most of the next hour about thalamus-level interfacing. The next morning, Corrigan extended his leave by a few days and caught a noon flight to San Francisco. He and Hans spent the rest of the afternoon talking in Hans’s lab at Stanford, and afterward into the early hours at Hans’s apartment, going through research notes and generating reams of charts and diagrams.

On returning to Pittsburgh, Corrigan went straight over to see Jason Pinder.

CHAPTER TWENTY-FIVE

Corrigan’s manner had changed since his last interview with Pinder. Although it had never come close to anything that could be called servile, common sense had always caused him to hold his opinions unless they were asked for, and then to couch them with a restraint appropriate to Pinder’s position. Now, however, the words poured forth as from an inspired evangelist. Pinder, aware that Corrigan was neither naive nor new to the business, listened with intrigued curiosity.

“Before the company leaps into putting up a lot of money and committing itself for years ahead, it ought to ask one last time what it stands to get in return for the investment,” Corrigan said. “When you sit down and analyze it, all that COSMOS is really promising is a more sophisticated version of what we’ve already got in the lab down there: a full-sensory interface. The only difference is that EVIE uses VIV for its vision and voice, whereas COSMOS will shift everything to the thalamus. But essentially it’s still the same thing. And that same thing is what the people from Feller and Faber told us they didn’t want—what Borth described as a ‘toy.’ What they do want, and what there’s still a huge market for out there if someone can come up with a way to achieve it, the thing that the industry has been after for decades, is
true AI
.” Corrigan drew a long breath as he came to the point that he was preparing to stake his future on. “Well, I think that I can deliver it.”

They both knew enough of what Corrigan was talking about to make questions unnecessary. All he needed was a cue. Pinder nodded. “Go on, Joe. How?”

Corrigan moistened his lips. “The top-down, analytical approach doesn’t work. Everyone in the field agrees. The only way it’s going to happen is by getting some kind of initially simple system to evolve.”

“Which has been tried in enough places too,” Pinder observed. “And the results have all been equally modest, to say the least.”

“Agreed. But they’ve all been tries at equipping computers with sensory apparatuses like TV cameras, arms, legs, and wheels, and letting them loose to explore some kind of environment. But you don’t realize how good biological nervous systems are until you try copying them. They were shaped by a billion years of evolution to interact with the real world. Computers weren’t.”

Which exhausted what everyone in the trade knew were the two acknowledged theoretical approaches. “So are you saying you know another way?” Pinder asked.

“Yes, I think so.”

“What?”

“Computers do interact extremely well with their own, internal worlds. . . . So what you do is, invert the conventional approach.” Corrigan spread his hands. “If training a machine intelligence in our world isn’t effective, let’s try doing it the other way around: by going into its world and doing it there.”

Pinder frowned. “Sorry, I’m not quite with you, Joe. Doing what, exactly? Where?”

“People interfacing via EVIE interact with a machine-created version of the real world through the surrogates that they control. But the machine could also put pseudopeople of its own in there too—‘animations.’ You design the system to be goal-directed to make the behavior of its animations converge to that of the real-people surrogates.”

Pinder sat back, seeing the implication at once and staring at Corrigan thoughtfully. “So its success would be measured through a kind of Turing test,” he said.

“Yes, exactly.”

“This is certainly a new one on me, Joe. I’ve never heard the like of it.”

“What do you think?”

“It’s intriguing.”

Corrigan could see that he was making an impression and pursued his point further. “The system wouldn’t need to know
why
the individuals that it was trying to imitate were doing whatever they did. Its brief would be simply to make its animations behave similarly, which it could accomplish from external observables. And that’s what’s different about this approach. In the past, we’ve always tried to press into service existing processing methods and associative structures—tools that were developed for other purposes. Well, very possibly they’re inherently unsuitable for this kind of job and can never work. But the way I’m talking about, the system will be free to create its own organization of associations and linkages in a way that’s appropriate to its goals.”

“Information-processing architecture is appropriate to what information-processing systems do. Whatever it is that has evolved inside cerebral cortexes is appropriate to what cerebral cortexes do,” Pinder summarized.

“That’s it. And we don’t need to know in advance what the final organization will be, any more than the first protoplasm needed to know the wiring for a mammalian brain. The system would learn the way children do: by trying to imitate ‘adults’ who already understand the way the world works, and making its own connections and associations accordingly.

“And we’ve got all the pieces needed to do it. Pinocchio provides the basics of a suitable vehicle for driving both the surrogates and the animations. EVIE, with the all-neural package that we’re talking about for COSMOS, gives us a mechanism for coupling in the surrogates. A multitasking expansion of Jenny Leddell’s Perseus system from MIT could drive the animations.”

Corrigan judged this a good place to stop at for a response, and waited. Pinder stroked his chin and stared down at the desk. What Corrigan was proposing was clear enough. He was searching for the flaws. Finally he looked up.

“A world to support that kind of evolution needs to be context-rich,” he said, meaning the degree of detail and its variability that the system would have to support. “The look-ahead for sudden context changes and recomputing SDVs still hasn’t been solved satisfactorily. And it would get a hell of a lot worse with this.”

It was an objection that Corrigan had expected. Now he could offer a radical departure from anything that had been considered so far. “COSMOS only gives us a bit sooner what EVIE would have led to anyway, eventually,” he said again. “But why get involved with the primary sensory system at all? If we are set on going straight to the thalamus, we can take advantage of new effects that operate beyond that level, that will crack that whole set of problems.”

Pinder looked surprised. “Effects? What effects are you talking about?” he asked.

“When I was in California last month, it wasn’t just for a romantic interlude and to get married,” Corrigan replied. “I also wanted to update myself on some work going on out there that I’d been following.” Not quite true, but it sounded better that way. “A group at Stanford is deep-coupling to the thalamus too. One of the people involved is called Hans Groener—I worked with him at MIT. His particular angle is dream research.”

“So how does it affect us?”

“Input compression. One of the things they’ve learned to do is to use a high-level code to activate percepts already stored in the nervous system. I think it could solve the details problem.”

“Dreams?” Pinder repeated. He thought about it and frowned. “But wouldn’t that make it all subjective? Everyone would experience their own world.”

“To some degree, maybe. But apparently there’s a commonality to the coding that has surprised everyone. So, yes, in a sense the participants would be experiencing what’s partly an induced dream; but—down to any level of detail likely to matter, anyway—the same dream. So the contextual environment would be much richer than anything we’ve’ ever contemplated before—and getting better all the time. The environment and the animations would stimulate each other into coevolving: one of the most powerful evolutionary mechanisms there is.”

Pinder looked as if he wanted to believe it. But there was one more reality to be faced. “Children need years to grow up,” he pointed out. “We don’t have years. What prompted the decision to go for COSMOS was that it gives us something to go for now.”

And that was that last thing that Corrigan had been waiting for. He nodded. “Yes, I know. And that’s where the other interesting thing that Hans’s people have stumbled on comes in. You know how it is when you dream—sometimes you find that what seemed to last hours all took place in a few seconds while you were waking up? Well, it seems that the effect can be achieved artificially when you go in above the primary sensory level.”

“Artificially?” Pinder’s eyebrows shot upward. “What are you saying? That it’s possible to accelerate interaction rates?”

Corrigan nodded. “Exactly that. Time in the simulated world could run faster. So you wouldn’t have to wait years for your child to grow up.”

“What kind of an acceleration are we talking about?” Pinder asked, now definitely interested.

“Somewhere in the hundreds, probably. That means that the equivalent of years of growing up would take a few weeks of machine time.” Corrigan sat back and extended a hand, palm upward, like someone offering the world. “There it is—all the ingredients for a true AI. And you could have it in as much time as we’re talking about now for COSMOS—which the customer says is just a toy.”

Pinder put the proposal to Ken Endelmyer, the CLC president, later that week, with the endorsement that in his opinion it was worth looking into seriously. Certainly, it was bold and vigorous in concept—maybe just what the whole field needed. A high risk, yes; but the potential rewards were huge, too, as they well knew. Endelmyer called in Therese Loel for an opinion. She was as intrigued as Pinder and agreed that there might be something in it. She also thought that the potential return from COSMOS was paltry compared to the market that this could open up. Endelmyer put the prospect, along with tentative estimates of what it would take to make the project fly, to the Board. Visions grew of this being pushed as the lead corporate research project, and it became a major funding issue. A month after Corrigan’s talk with Pinder, orders came down from corporate headquarters to put the present plans for COSMOS on hold.