“Did he figure that out himself?” Evelyn whispered. Jenny looked across at her and read the expression on her face.
“It’s getting to you, isn’t it,” she said understandingly. “It’s okay. Don’t worry. A lot of people are affected like that. Maybe that’s enough for the first time.” She killed the screen.
Evelyn was still not at ease. “Is he still active in there, while the screen’s off? Or does he go into a suspended state until you switch it on again?” she asked.
“It’s just an illusion,” Jenny said. She looked at Corrigan. “Shouldn’t you be getting along, anyhow? Didn’t you say something about wanting to catch Marvin before he leaves?”
* * *
They met Minsky in a staff cafeteria on the second floor, where he was grabbing coffee and a sandwich before dashing off to keep an appointment elsewhere on the campus. Tall, smooth-domed, continually observing the world through thick-rimmed spectacles but never quite able to take it seriously in its entirety, he was one of the lab’s original founders. Corrigan had known him sporadically in his time at MIT and was pleased that their schedules had enabled a meeting during this quick visit.
Minsky, it turned out, had returned from Ireland himself recently, where he had been partly vacationing and partly checking the state of contemporary computing developments at Trinity and University colleges. His experiences from a drive north to Ulster, where trouble was still going on with the British administration, had left him less than impressed, however.
“Why are they still fighting each other up there?” he grumbled to Corrigan. “Why don’t they study mathematics, or something else that would give them better things to do?”
“Oh, that’s nothing to do with us,” Corrigan replied. “It’s another country up there. I’m from the Republic, remember.”
Minsky pulled a face. “I’m not sure I noticed much of a difference. Down there, if you’re an American and don’t know the price of anything, you’re fair game.”
“The lads have to make a living,” Corrigan said unapologetically, refusing to be provoked.
“You mean it isn’t true, what the song says?” Evelyn put in. ” `When Irish eyes are smiling . . .’?”
“You’ve probably been ripped off,” Minsky completed with a snort. Corrigan laughed. Minsky glanced at his watch. “Anyway, I have to dash in a few minutes. So I gather you’ve been to see Jenny Leddel.”
“Perseus is coming along nicely,” Corrigan said. “Evelyn got a bit spooked, though.”
“There was something eerie about it.” She shivered and shook her head.
Minsky smiled. “Yes. It gets a lot of people like that. It makes them wonder if we’re inside someone else’s AI experiment in the same kind of way.”
“That’s exactly what I was thinking,” Evelyn said, astonished. Minsky’s smile widened.
“The approach seems to be working,” Corrigan observed. He glanced at Evelyn. “Setting it up with the. potential to learn, and then letting it interact with an environment.”
“Jenny should have let Perseus start out as more of an infant,” Minsky commented. “There are still too many defined attributes. Instead of telling him what a sword is for, let him wave one about and hit things with it, and find out for himself. That way, he might even discover things that programmers never think to include—such as, that they make good back-scratchers.”
Corrigan related the episode of Perseus and the lamp. Minsky nodded emphatically.
“Which makes my point. He should have been allowed to read picture books and fairy tales. Then he would have been familiar with genies and known what to do.”
Evelyn was about to ask if he meant literally exposing a computer to the processes that a child goes through, say, by equipping it with appendages of some kind to manipulate things, but Minsky preempted her. Corrigan was used to his sometimes disorienting habit of getting people out of step in a conversation by answering questions before they were asked.
“Computers aren’t very good at interfacing with the real world and extracting the information they need. We have the advantage of this enormous knowledge-base that we call `common sense,’ which enables us to make subtle, context-based connections. That’s what makes people so good at things like comprehending metaphors: we’re wired to see quickly what matters and what doesn’t. Recognizing faces is another good example.” He waved a hand as he collected together the paper plate, coffee cup, and remains of his sandwich. “Computers are better at tasks that don’t require any deep familiarity with what’s out there—ones that can be dealt with in relative isolation, algorithmically.”