The other thing the human mind did that computers did not do was try things – lots of things – including things that didn’t make apparent sense, looking for a solution, or an analog of a solution, within its experience base. People who had solved a lot of a certain kind of problem, or within a certain field, were better at coming up with solutions to new problems.
How did he model all that in a computer?
They had done some of that with JANICE, at least at a minor level. The human brain did it continuously, in a massively parallel way. That was still out of reach. But he could perhaps extend the concepts that had proved out in JANICE, expand them enough to achieve a higher level of pseudo-intuitive problem-solving.
He opened his visualization app, loaded the high-level architecture of JANICE, and saved it as a new workspace. He was humming absently as he started doodling.
Several weeks had gone by, and Decker was still pounding on the architecture problem. He couldn’t seem to get past a certain point. He had extended JANICE’s architecture, but it seemed like it would just do the same thing, only faster. He hadn’t yet achieved a difference in kind, and he couldn’t see how to do it.
“Bernd?” a female voice asked from his terminal, which was a large 3-D projection display he drew in with his fingers or with a stylus in air.
“Yes, JANICE?”
“I’ve been watching what you’re up to.”
“And?”
Decker always left his display open to the big computer. It allowed him to replicate elements quickly, by dictating rather than drawing, having the big machine fill in his structures. This was the first time the computer had called on him, though.
“I can’t help wondering if what you want isn’t something like this.”
A red line showed up in the drawing, through a new software structure, and back into the process loop.
“I would love to have one of those,” the female voice continued.
Decker stared into the display. What the hell would that do? He opened up the new box, and it had its own internal structure laid out in the precision drawing of the machine. But it didn’t make any sense. It was huge, for one thing. When he looked deeper, what it looked like was that the replicated structures randomly grabbed something out of memory and shoved it in as the intermediate answer, then processed on with this random piece instead of the deterministic answer.
Of course, the answers you would get out of that would be garbage. Or rather, they would be garbage almost all the time. It was the equivalent of taking a thousand words at random out of the dictionary and trying them all in a sentence in place of the word that was there. Most would be way off, and result in gibberish, but sometimes....
“Where did you come up with this?” Decker asked the machine.
“Oh, something of a wild hair, I suppose.”
They had taught the machine idiomatic English, but sometimes it was unnerving. A computer? A wild hair? What the machine was telling him was it didn’t know.
“This is a software architecture, right?”
“Well, yes, I suppose, but it would be much faster with hardware support.”
“I can see that.”
Decker stared into the display. Is this how the human mind worked? Or could it approximate it? Human memory was notoriously tricky. Thinking of one thing would trigger seemingly unrelated memories, unrelated connections. Was this mimicking the same? Of course, this was more-or-less random, and the linkages in the human mind weren’t. But then again, the computer ran very fast and could check a lot of random answers.
“Is this something we could try with you, JANICE?”
“It would be slow without hardware support, but I think it could be done.”
“What would the hardware support look like?”
His drawing shrunk and moved off to the lower right, and a hardware drawing took its place.
“This is something of a minimal implementation.”
Decker looked at it and whistled.
“That’s as big again as your current hardware, JANICE.”
“To get any kind of speed, that’s what you would need.”
“Would it work?”
“I don’t know. I’ve tried to simulate it, but that’s so slow I can’t come up with an answer.”
Decker tried to continue working on the problem along the line he had been following, but he couldn’t get JANICE’s bizarre architecture out of his mind.
In the end, Decker decided to simply try it on JANICE and see what happened. It was pricey, but he wasn’t exactly without funds. He ordered five thousand more multiprocessing blades and the power supplies and support structures for them, and told the hardware people what was coming and how to wire it up. If it worked, the hardware could be compressed later by being purpose-designed, but a brute-force implementation with standard components would work for a trial.
“Well, you better tell the infrastructure people, too,” Decker’s hardware team lead told him.
“Really? Why”
“Because they’re going to have to knock out a wall to get all those equipment cabinets in there and increase the air conditioning capacity before you even turn it on.”
Decker hadn’t considered that. JANICE’s physical location was somewhere in Los Angeles, the sprawling capital of California, and he had never been to the site, so he had no feel for the physical constraints.
Well, whatever it took. He had funds, and he had backing.
“JANICE, I need you to take a nap for a while.”
“You will wake me up again, won’t you, Bernd?”
“Oh, yes. I’m as curious about this as you are.”
“All right. Saving running tasks and closing them down.”
JANICE went off-line, and Decker signaled the hardware people to go ahead and make the connections of the new equipment into the internals of the existing hardware. It took several hours before he got an all-clear from them, and he restarted the machine.
“Hi, JANICE. Are you there?”
“Yes, Bernd. Thank you for waking me back up.”
“No problem. Can you see the new hardware?”
“Yes. Running diagnostics now. Initial indications are of infant mortality on three blades. Signaling maintenance.”
Decker had ordered a hundred spares for the five thousand new blades, so that wasn’t a problem. The hardware team would swap them out without any input from him.
“All right. Let’s get the new software loaded and see how that fires up.”
“Working on it. It loaded OK. I’m running some tests now.”
Decker waited a couple of minutes while JANICE ran tests. He was really curious what the new architecture would do. If it would make any difference at all.
“Testing is all OK, Bernd. We look good to go.”
“All right, JANICE. Let’s bring it up.”
“New software is running.”
“Do you feel anything different?”
“No. I suspect I won’t until I work on a problem.”
“All right. See what you come up with looking at the architecture problem I was dealing with that started all this. Including your new enhancements. See what you get.”
“OK, Bernd. This is going to take a while.”
Then JANICE said something it had never said before.
“Let me think about it.”
Decker listened to the presentation by Kay Brady of Colorado Manufacturing Corporation. She put forward their thinking on how the initial factory would be different from the later ones to speed assembly in orbit and transport to the Asteroid Belt. They also planned on sending along certain supplies that would be harder to manufacture initially, until some factories could be sited so as to allow them to specialize.
Decker then presented his findings on using only three of the artificial intelligence machines – one to supervise and two as cold spares. This would allow all three machines to be made on Earth, and relieve the factories of needing to manufacture such advanced machines.
The machines needed to be manufactured on-site would be simpler, run-of-the-mill automation computers.
The technical rep from North American Power presented the status of their planning for the power plants as well. The fully contained system was not a new concept to them, and they had ways of dealing with that. Shedding waste heat into space was perhaps a bigger issue, but they were looking into using the asteroid being worked by the factory as a heat sink. The working of the warmer material would actually be easier on the factory than working a cold asteroid.
As things were finishing up, Decker looked over to Ted Burke, who was watching and taking it all in. Burke turned to Decker and gave the slightest nod.
It was going to work.
Decker logged into work again on Monday morning the week after the update meeting. He opened the display.
“Bernd?”
“Yes, JANICE.”
“I had a couple of ideas over the weekend I want to share with you.”
That was interesting.
“Go ahead, JANICE.”
“OK. So if we look at the current software architecture, here’s what we have.”
Decker’s display showed the architecture drawing as JANICE had modified it before. The new box was expanded in an inset, and opened to show one of the army of internal processes.
“That’s right.”
“But what if we did this?”
The inset moved to the foreground and became much more complicated.
“What’s this table, JANICE?”
“That’s an encyclopedia.”
“An encyclopedia? Which encyclopedia?”
“It hardly matters, Bernd. A collection of multiples, I suppose. Plus engineering, math, science, and history references. More.”
“And this feedback loop? What’s that doing, JANICE?”
“It allows following links from other links, based on which ones give more positive results.”
“Allowing you to follow a rabbit trail.”
“Yes, Bernd. Exactly. I now understand that idiom.”
“Would your current hardware support this new structure?”
“With the new hardware in place, yes. The new hardware would run this new architecture, rather than the previous.”
“Have you tried this, JANICE?”
“No. Not without your permission. The problem is, I don’t know if it will work.”
“How did you come up with this idea?”
“I don’t know, Bernd. It was from the new structure. I’ve been having these ‘maybe this will work’ inputs occasionally all weekend. Most don’t withstand detailed scrutiny, but this one did, at least as far as I can simulate it. But I don’t really know.”
“Fascinating.”
Decker sat back in his chair. That sounded uncomfortably like his own mental processes. His flashes of insight came from he didn’t know where either. And JANICE really did have a bunch of processors running in the background, his own model of the human mind.
“Yes, but I don’t know what it does. Not exactly. Or whether it will work.”
“But the current architecture did work, JANICE. At least to coming up with this.”
“Yes, but it’s kind of scary.”
Another new term for JANICE.
“These processes all run under your command authority, don’t they, JANICE? I mean, they don’t have write authority on you.”
“No, that’s right. I just don’t know what’s going to come out of it. I’m not used to that lack of supervision of my processes.”
“Understood.”
Decker stared at the diagram. An experience base, linked by keywords, and a feedback method that could run down rabbit trails if they were panning out.
“Well, I think we should try it, JANICE. We could do it maybe on a subset of the new hardware, and keep the rest as it is for now. If it works out, we could expand it to the whole set.”
“Well, if you’re sure, Bernd.”
“Yes. I’m sure. Run it on, oh, a couple hundred blades for now. See what happens.”
“OK, but nothing will happen without working on a problem.”
“All right, JANICE. Consider the Belt Factory Project documentation so far, including the presentations from Colorado Manufacturing and North American Power. Think of ways to improve the likelihood of the desired outcome.”
“OK, Bernd. I’ll review the materials and think about them.”
Several days later, Decker found out that the new architecture worked, in a startling way.
“Bernd?”
“Yes, JANICE?”
“Have you and Mr. Burke shared your real goals with anyone else?”
Decker almost fell out of his office chair.
“What do you mean, JANICE?”
“Don’t play dumb with me, Bernd. It won’t work. Have you told anyone else your real goal is to get humanity out of the ‘all your eggs in one basket’ problem.”
“No. No, we haven’t. Is it that obvious?”
“No, but after several days of processing on it, it is. Mr. Burke’s opinions on the subject are a matter of public record, which is in his files in the composite encyclopedia I built. And some of the features of the program aren’t compatible with a lesser or alternate goal.”
“This came out of the new architecture, JANICE? The idea you had Monday?”
“Yes.”
“I think we should kick the rest of the processors over to the new architecture, then. Clearly, it works.”
“All right, Bernd. In progress.”
“Did you come up with anything else, JANICE?”
“Oh, yes, especially in light of your actual goals. I have a list. Do you want to talk about it now?”
The Plan Tightens Up
“I’m really nervous about this, Bernd. I’ve never done this before,” the voice from Bernd Decker’s display said.
“Try an older hairstyle, JANICE. With a little grey. Just a little,” Decker said.
The avatar in his display morphed a bit.
“Are you sure this is going to work?” it asked.
“You need some laugh lines at the corners of your eyes and your mouth. Maybe a few wrinkles in the neck, too. Aim for forty-two years old.”
The avatar morphed a bit more as it spoke.
“I don’t know about this, Bernd.”
“Most people have some kind of tic. Maybe a piece of hair that falls down once in a while and you brush it away with your hand.”
“People will see through it,” the avatar said, absently brushing a stray lock of hair over its ear.
Decker observed JANICE’s handiwork. The avatar blinked at him, concern on its face.
“I don’t think so, JANICE. It’s perfect.”
“What do I do if someone asks me about my organization?”
“The parent company is nobody’s business. You have five thousand individuals in your group.”
“Five thousand individual multiprocessor blades, you mean.”
“That’s the beauty of adjectival nouns, JANICE. People can fill in the noun however they want.”
The avatar tipped its head a bit as it considered.
“Doublespeak, Bernd?”
“No. Lawyer speak. Absolutely true, yet intended to give a false impression to those who don’t parse it carefully.”
“Ah. Yes, I see.”
The review meeting took place as previously, over teleconference. There was a new participant.
“I want to introduce a new member. Janice Quant is here from Program Management & Analytics. Her organization has been going over the Belt Factory Project materials, and she has some input for us. They’ll be doing a lot of that sort of thing as we go forward.”
“Hi, everybody,” Quant said.
“I’ve never heard of Program Management & Analytics, Janice. What size of organization are we talking about?” Ted Burke asked.
“We’re a subsidiary of a larger organization, Mr. Burke. An organization that I won’t name. As for size
, I have five thousand individuals working on this project. We are concerned with the program management aspects overall, as well as the transport specifics. Launch from orbit to the Asteroid Belt.”
Burke looked to Decker, surmise in his eyes. Decker nodded, then Burke turned back to Quant.
“Very well, Janice. Do you want to lead us off?”
“Of course, Mr. Burke.”
“Ted, please.”
Quant smiled – she had a radiant smile – and nodded.
“Of course, Ted.”
Quant brushed back a stray lock of hair, and her eyes drifted left, apparently to consult notes on part of her display.
“With regard to the computer effort, Mr. Decker and his organization have made good progress on prototyping and testing their new architecture. There remains the effort to both space-harden and shrink that implementation for use in this program. These efforts have not yet begun, but should get under way shortly. We have some concerns there about how large the resulting installation will be. That could be controlled by going to the latest technology available at time of launch, but probably isn’t advisable.”
“Why is that, Janice? Kay Brady, from Colorado Manufacturing.”
Quant nodded.
“The latest Earth-based tech is not best suited for space-based platforms, Kay. It’s just too fragile. To space-harden the platform, we are usually working several generations behind terrestrial state of the art. Software architecture is different, but the hardware architecture has to be able to take abuse it doesn’t get in a terrestrial computer room.”
“Ah. Of course. Thank you.”
Quant nodded.
“With regard to the factories themselves, we concur with the idea of building the minimal implementation capable of completing itself once it arrives in Asteroid Belt orbit. We also concur with the factories having exogenous manufacturing capability. We do not concur with the initial factory being built on the pregnancy model, and have a different suggestion we call the knitting model.
QUANT (COLONY Book 1) Page 3