Okay, he said, lighting a cigarette. And the next step?

The next wasn't really a technical step forward in telefactoring. It was an economic shift. The purse-strings were loosened and we could afford to send men out. We landed them where we could land them, and in many of the places where we could not, we sent down the telefactors and orbited the men again. Like in the old days. The time-lag problem was removed because the operator was on top of things once more. If anything, you can look at it as a reversion to earlier methods. It is what we still often do, though, and it works.

He shook his head.

You left something out between the computers and the bigger budget. I shrugged.

A number of things were tried during that period, but none of them proved as effective as what we already had going in the human-computer partnership with the telefactors.

There was one project, he said, which attempted to get around the time-lag troubles by sending the computer along with the telefactor as part of the package. Only the computer wasn't exactly a computer and the telefactor wasn't exactly a telefactor. Do you know which one I am referring to?

I lit a cigarette of my own while I thought about it, then, I think you are talking about the Hangman, I said.

That's right and this is where I get lost. Can you tell me how it works?

Ultimately, it was a failure, I told him.

But it worked at first.

Apparently. But only on the easy stuff, on To. It conked out later and had to be written off as a failure, albeit a noble one. The venture was overly ambitious from the very beginning. What seems to have happened was that the people in charge had the opportunity to combine vanguard projects, stuff that was still under investigation and stuff that was extremely new. In theory, it all seemed to dovetail so beautifully that they yielded to the temptation and incorporated too much. It started out well, but it fell apart later.

But what all was involved in the thing?

Lord! What wasn't? The computer that wasn't exactly a computer ... Okay, well start there. Last century, three engineers at the University of Wisconsin, Nordman, Parmentier, and Scott, developed a device known as a superconductive tunnel-junction neuristor. Two tiny strips of metal with a thin insulating layer between. Supercool it and it passed electrical impulses without resistance. Surround it with magnetized material and pack a mass of them together, billions, and what have you got? He shook his head.

Well, for one thing you've got an impossible situation to schematize when considering all the paths and interconnections that may be formed. There is an obvious similarity to the structure of the brain. So, they theorized, you don't even attempt to hook up such a device. You pulse in data and let it establish its own preferential pathways, by means of the magnetic material's becoming increasingly magnetized each time the current passes through it, thus cutting the resistance. The material establishes its own routes in a fashion analogous to the functioning of the brain when it is learning something.

In the case of the Hangman, they used a setup very similar to this and they were able to pack over ten billion neuristor-type cells into a very small area, around a cubic foot. They aimed for that magic figure because that is approximately the number of nerve cells in the human brain. That is what I meant when I said that it wasn't really a computer. They were actually working in the area of artificial intelligence, no matter what they called it.

If the thing had its own brain, computer or quasihuman, then it was a robot rather than a telefactor, right?

Yes and no and maybe, I said. It was operated as a telefactor device here on Earth, on the ocean floor, in the desert, in mountainous country, as part of its programming. I suppose you could also call that its apprenticeship, or kindergarten. Perhaps that is even more appropriate. It was being shown how to explore in difficult environments and to report back. Once it mastered this, then theoretically they could hang it out there in the sky without a control loop and let it report its own findings.

At that point would it be considered a robot?

A robot is a machine which carries out certain operations in accordance with a program of instructions. The Hangman made its awn decisions, you see. And I suspect that by trying to produce something that close to the human brain in structure and function, the seemingly inevitable randomness of its model got included in. It wasn't just a machine following a program. It was too complex. That was probably what broke it down.

Don chuckled.

Inevitable free will?

No. As I said, they had thrown too many things into one bag. Everybody and his brother with a pet project that might be fitted in seemed a supersalesman that season. For example, the psychophysics boys had a gimmick they wanted to try on it, and it got used. Ostensibly, The Hangman was a communications device. Actually, they were concerned as to whether the thing was truly sentient.

Was it?

Apparently so, in a limited fashion. What they had come up with, to be made part of the initial telefactor loop, was a device which set up a weak induction field in the brain of the operator. The machine received and amplified the patterns of electrical activity being conducted in the Hangman's, might well call it 'brain', then passed them through a complex modulator and pulsed them into the induction field in the operator's head ... I am out of my area now and into that of Weber and Fechner, but a neuron has a threshold at which it will fire, and below which it will not. There are some forty thousand neurons packed together in a square millimeter of the cerebral cortex, in such a fashion that each one has several hundred synaptic connections with others about it. At any given moment, some of them may be way below the firing threshold while others are in a condition Sir John Eccles once referred to as 'critically poised', ready to fire. If just one is pushed over the threshold, it can affect the discharge of hundreds of thousands of others within twenty milliseconds. The pulsating field was to provide such a push in a sufficiently selective fashion to give the operator an idea as to what was going on in the Hangman's brain. And vice versa. The Hangman was to have its own built-in version of the same thing. It was also thought that this might serve to humanize it somewhat, so that it would better appreciate the significance of its work, to instill something like loyalty, you might say.

Do you think this could have contributed to its later breakdown?

Possibly. How can you say in a one-of-a-kind situation like this? K you want a guess, I'd say, 'Yes.' But its just a guess.

Uh-huh, he said, and what were its physical capabilities?

Anthropomorphic design, I said, both because it was originally telefactored and because of the psychological reasoning I just mentioned. It could pilot its own small vessel. No need for a life-support system, of course. Both it and the vessel were powered by fusion units, so that fuel was no real problem. Self-repairing. Capable of performing a great variety of sophisticated tests and measurements, of making observations, completing reports, learning new material, broadcasting its findings back here. Capable of surviving just about anywhere. In fact, it required less energy on the outer planets, less work for the refrigeration units, to maintain that supercooled brain in its midsection.

How strong was it?

I don't recall all the specs. Maybe a dozen times as strong as a man, in things like lifting and pushing.

It explored Io for us and started in on Europa.

Yes.

Then it began behaving erratically, just when we thought it had really learned its job.

That sounds right, I said.