Выбрать главу

So they descended, by elevator instead of scuba gear. The Dentrassi brothers had opened a new restaurant underwater, in a geodesic dome on the shallow seafloor. Their prices were too high, geared for tourists who wanted the novelty, but Tess had a standing discount. She asked for a “corner table” just to tease the waiter, then went with her guests to a booth at the round dome’s edge.

Clark said, “Where is he?”

Just then, Zephyr arrived and took a chair to sit beside the booth. “Good afternoon. Tess has done the gushing on my behalf, so I’ll skip it.”

Emi leaned toward him, fascinated. “Some artistry went into your body design. It’s very sleek.”

“Thanks. I’ve just gotten some more background info on you; you’re some sort of artist and Clark there is a businessman?”

Clark nodded. “I focus on hardware. Industrial control systems.”

Tess said, “I’d send you to meet Valerie, but she’s busy. Captain Fox’s son is still in alpha development, not ready for public release.”

Zephyr silently sent, I’m here, to quell the jealousy Tess still felt toward Valerie. She was wrong to keep pining for Captain Fox, when she had a soulmate of her own. She didn’t need to say anything back to Zephyr; he knew exactly how she felt.

Zephyr had a glass of water ceremonially placed in front of him, for appearance’s sake. After several minutes of pleasantries he said, “Well?”

Tess shot him a look and a thought of that’s a bit rude. Aloud she said, “We take it there’s something more to this meeting than asking each other for autographs.”

Emi had been doodling on a napkin, though when she pulled her hand away Tess marveled at the elaborate flower design she’d conjured from nothing. Emi said, “The three of us are teaming up to experiment with game design and AI. Do you want to get involved?”

“In making games?” asked Tess, surprised. She was a little busy these days, what with being part-owner of a mad science company. “Westwind Transhuman Designs” didn’t do games, though it made robots and was dabbling with gengineering and other technology.

Researcher Alain spoke quietly. “In developing an AI-powered game, one that can play with many people at once.”

“Most games already have that; the AIs are just morons meant to die and make the player feel cool and powerful.”

“We’d like to go a step farther. You’re already aware of my little creations that can carry a conversation and control a virtual body in a simple game world.” Those were part of his published work. “People assume that the next step is to take that AI technology from a simulation into the real world, and create a direct competitor for the Hayflick series.” Alain nodded toward Zephyr.

Zephyr’s eyes flickered between orange and green, for him a sign of guarded, wary thoughts. Tess picked up only snatches of What’s their scheme here? Zephyr said, “I understand from your phrasing that you have a different idea.”

Emi explained, “Zephyr, you’ve found your own understanding of reality and human culture by occupying a humanoid body. You also have contact with this regional network for exchanging thoughts and ideas. We’d like your expertise for creating a non-embodied AI system that learns from its players.”

“Like a surveillance system?” asked Tess.

Clark waved off that idea. “No. We’re… not fans of those.”

Alain leaned forward. “There is a certain urgency to our work, frivolous as it may seem.”

Zephyr spoke rapid-fire to Tess, who figured out quickly what he was getting at. The robot had temporarily disconnected the two of them from any risk of sharing with anyone else in their network, so that he could say to her, Reading between the lines, they’re going for the Holy Grail.

Tess answered him, And they’re scared of something.

Self-improving AI was technically possible; Zephyr himself was the poster-bot for that. What people usually meant by the term, though, wasn’t just that a machine would read books and form opinions. Instead, they meant a system that rapidly altered its own code. The craziest predictions said that some autistic genius would leave his AI running overnight and foom!, it’d ascend to godlike superintelligence and probably destroy the world, due to having vast resources and no common sense. The Castor network wasn’t on that level at all, nor was Zephyr.

But this trio of game-makers seemed to have a plan for gathering gobs of data on human interaction, and DeLune had already built code with a limited form of self-editing and growth. They might create a new type of AI that could rapidly upgrade itself while being somewhat grounded in sanity. Or just one that was really, really good at manipulating people.

Tess and Zephyr made all these connections in about ten seconds while they sat there slack-jawed. Then came the other piece of the puzzle: By ‘urgency’, they’re hinting that there’s less ethical competition out there.

Tess composed herself and said, “One to ten: how screwed do you think the human race is right now?”

Clark chuckled mirthlessly. “Due to AI specifically, maybe a three. We’re aware of several, let’s say, well-funded organizations that want to invent an AI that goes well beyond Zephyr’s abilities. Impressive as they are. And of course they have access to the same research of Alain’s that you’ve seen.”

Alain muttered, “Should never have published.”

Emi added, “You’re well aware of what can hold an AI project back. We’re not eager to help the competition re-learn the lessons that you did, or to help them find a workaround.”

Tess leaned back in her seat. “So. You want to produce a new AI system of a type that everyone says is super dangerous, on the theory that someone else might do it first?”

Zephyr said, “And label it as a game.”

Clark said, “You get it. You could help push our project along faster, and safer.”

Tess had been involved in some strange activities over the last few years, from running away from home, to building a town out of nothing. This idea, though, was far out by her standards. She tapped the table. “Zephyr and I are tied to a specific location, a specific set of senses and goals. We can’t save the world from some hypothetical god-tier AI, but that’s not realistic anyway. You get ideas and success from trying to solve problems here and now, not from finding some perfect theory that will Fix Everything.”

Clark said, “But we can find a system that will help address everyone’s problems at once. If we just find the right code base and iterate—”

Zephyr’s laugh was a synthetic echo of Tess’ voice. They started to speak the same words, but Zephyr deferred to Tess. She said, “First piece of advice: nope, finding the one true idea isn’t going to save you. Second thing: no, you’re not going to solve everybody’s problems; hell, you’ll probably invent new ones.”

Zephyr added, “Not that we think you should stop. Just… try for something simpler, and build from there.”

Emi said, “Like what?”

Tess shrugged. “Can your AI system figure out how to help the players have fun?”

“You think that’s a straightforward task?”

“No, but if you can’t solve it, you’re not ready for bigger ones.”

“Fun, huh,” said Clark, playing with his silverware. “It doesn’t mean the same thing to everyone.”

Tess said, “Exactly.” If these three ever got anywhere beyond Alain’s existing research, it’d be through finding what each individual wanted, not what “people” wanted.

When the meal was over, the guests stood, and shook hands again with Tess and Zephyr. “You’ve given us a lot to think about,” Alain said. “I think we’ll be seeing each other again, as our work progresses. Until then, would you mind giving us a little tour on our way out?”