Harry Harrison - The Turing Option

“I can’t wait,” Shelly said. “Not that the food here is bad — but I would love a change.”

“How is the investigation going?” Brian asked. It was never far from his thoughts. Ben frowned into his beer.

“I wish I could bring some good news, but we seem to have hit a dead end. We have Alex Toth’s military record. He was an outstanding pilot, plenty of recommendations for that. But he is also a borderline alcoholic and a troublemaker. After the war they threw him out as fast as they could. No trace of him at the address he gave at the time. The FBI has found some records of his employment through his pilot’s license, kept up to date. But the man himself has vanished. The trail is ice cold. Dusty Rhodes’ story checks out. He was conned into it and then left to hang out and dry in the wind. There is absolutely no way to trace the money that was paid into his account.”

“What’s going to happen to Rhodes?” Shelly asked.

“Nothing now. The remaining money they gave him has been sequestered for the crime victims’ fund and he signed a complete statement of everything that happened, everything he did. He’ll keep his nose clean in the future or will be hit with a number of charges. We want to keep this thing as quiet as we can while the investigation is still in progress.”

Shelly nodded and turned to Brian. “You must bring me up to date. Did you ever get that B-brain to work?”

“Indeed I did, and sometimes it works amazingly well. But not often enough to trust very far. It keeps breaking down in fascinating and peculiar ways.”

“Still? I thought that using LAMA-5 made debugging easier.”

“It certainly does — but I think that this is more a problem of design. As you know, the B-brain is supposed to monitor the A-brain, make changes when needed to keep it out of various kinds of trouble. Theoretically this works best when the A-brain is unaware of what is happening. But it seems that as Robin’s A-brain became smarter it learned to detect that tampering — then tried to find ways to change things back. This ended up in a struggle for power as the two brains fought for control.”

“It sounds like human schizophrenia or multiple personalities!”

“Exactly so. Human insanity is mirrored in machine madness and vice versa. Why not? A malfunctioning brain will have the same symptoms from the same cause, machine or man.”

“It must be depressing, being set back by lunatic brains in a box.”

“Not really. In a way, it’s actually encouraging! Because, the more the robot’s foul-ups resemble human ones, the closer we are getting to humanlike machine intelligence.”

“If it is going that well — why are you so upset?”

“Is it obvious? Well, it’s probably because I’ve finally come to the end of the notes we retrieved. I’ve worked through just about everything that those notes described. So much so that now I am swimming out into uncharted seas.”

“Is there any rule that the AI in your lab must be the same as the one that was stolen?”

“Yes, pretty much so, except for some minor details. And the trouble is that it has so many bugs that I am afraid that we’re stuck on a local peak.”

“What do you mean?” Ben said.

“Just a simple analogy. Think of a scientific researcher as a blind mountain climber. He keeps climbing up the mountain and eventually reaches a peak and can climb no higher. But because he can’t see anything he has no way of knowing that he’s not at the top of the mountain at all. It is merely the peak of a local hill — a dead end. Success is then not possible — unless he goes back down the mountain again and looks for another path.”

“Makes sense,” Ben said. “Are you telling me that the AI you have just built — which is probably almost the same as the one that was stolen — may be stuck on a local peak of intelligence and not on some much higher summit?”

“I’m afraid that’s it.”

Ben yodeled happily. “But that is the best news ever!”

“Have you gone around the twist?”

“Think for a second. This means that whoever stole your old model must also be stuck in about the same way — but he won’t even know it. While you can go and perfect your machine. When that happens we’ll have it — and they won’t!”

As this sunk in a broad grin spread across Brian’s face. “Of course you’re right. This is the best news ever. Those crooks are stuck — while I’m going to push right ahead with the work.”

“Not at this moment you’re not — after lunch!” Shelly said, putting down her wineglass and pointing to the door. “Out. It’s after two and I’m starving. Eat first, talk later.”

After eating See Khrong Moo sam Rot — which despite its name was absolutely delicious — sweet, sour and salty spareribs — they even managed some custard steamed in pumpkin for dessert.

“I’ll never eat army chow again,” Brian groaned happily and rubbed his midriff.

“Tell that to the cook — make his day,” Shelly said. “That’s what I’m going to do.”

Lat Phroa took their praise as his due, nodding in agreement. “It was pretty good, wasn’t it? If the rest of the troops like it I’m going to work hard to get this kind of chow in the regular menu. If only for my own sake.”

Ben left them there and they walked off some of the lunch by strolling back to the lab.

“I’m enthusiastic — but apprehensive,” Brian said. “Swimming out into uncharted seas. Up until now I have been following the charts, my own notes — but they have just run out. It’s a little presumptuous of fourteen-year-old me to think that I can succeed where the twenty-four-year-old me pooped out.”

“Don’t be so sure. Dr. Snaresbrook maintains that you’re smarter now than ever before — your implants have given you some outstanding abilities. And furthermore, in the work you’ve done with Snaresbrook — analyzing your own brain — you’ve probably discovered more about yourself than a squad of psychologists ever could. It’s clear to me that you’re getting there, Brian. Bringing something new into the world.

“A truly humanlike machine intelligence.”

Ben found the message in his phone when he woke up. It was Brian’s voice.

“Ben — it’s four in the morning and we have it at last! The data in Robin was almost enough, and Dr. Snaresbrook finished the job by decoding some more material from my brain. It was an awful job, but we managed to get it done. So now, theoretically, Robin contains a copy of my superego and I’ve set the computer to reassembling all of Robin’s programs to try to integrate the old stuff with the new. Need some sleep. If you can make it please come to the lab after lunch for a demo. Over and out — and good night.”

“We’ve done it,” Brian said when they met in the laboratory. “The data already downloaded into Robin was almost enough. It was Dr. Snaresbrook who finished the job, adding what might be called a template, a downloaded copy of my superego. You could say that it was a copy of how the highest-level control functions of my brain operate. All memory that was not associated with control was stripped away until we had what we hoped would be a template of a functioning intelligence. Then came the big job of integrating these programs with the AI programs that were already running. This was not easy but we prevailed. But along the way we had some spectacular failures — some of which you already know about.”

“Like the lab wreck last week.”

“And the one on Tuesday. But that is all in the past. Sven is now a real pussycat.”

“Sven?”

“Really Robin number 7, after we found out that 6.9 couldn’t access all the memory we needed.”