of vision with the scene below. He also wore earphones connected with its audio pickup. I read the book he wrote later. He said that for long stretches of time he would forget the cabin, forget that he was at the boss end of a control loop, and actually feel as if he were stalking through that hellish landscape. I remember being very impressed by it, just being a kid, and I wanted a super-tiny one all my own, so that I could wade around in puddles picking fights with microorganisms."

"Why?"

"Because there weren't any dragons on Venus. Anyhow, that is a telefactoring device, a thing quite distinct from a robot."

"I'm still with you," he said, and "Now tell me the difference between the early telefactoring devices and the later ones."

I swallowed some coffee.

"It was a bit trickier with respect to the outer planets and their satellites," I said. "There, we did not have orbiting operators at first. Economics, and some unresolved technical problems. Mainly economics. At any rate, the devices were landed on the target worlds, but the operators stayed home. Because of this, there was of course a time lag in the transmissions along the control loop. It took a while to receive the on-site input, and then there was another time lapse before the response movements reached the telefactor. We attempted to compensate for this in two ways: the first was by the employment of a single wait-move, wait-move sequence; the second was more sophisticated and is actually the point where computers come into the picture in terms of participating in the control loop. It involved the setting up of models of known environmental factors, which were then enriched during the initial wait-move sequences. On this basis, the computer was then used to anticipate short-range developments. Finally, it could take over the loop and run it by a combination of 'predictor controls' and wait-move