Robot learning response
A robot could be repogrammed to produce new behaviours by learning in the same way humans do. A program that learned to tell new jokes would not repeat jokes the programmer had entered into its memory, but would invent jokes as humans do.
The Putnam argument
"Some of the arguments designed to show that Oscar could not be conscious may be easily exposed as bad arguments. Thus, the phonograph-record argument: a robot only 'plays' behavior in the sense that a phonograph record plays music. When we laugh at the joke of a robot, we are really appreciating the wit of the human programmer" (H. Putnam, 1964, p. 679).
"The first argument ignore the possibility of robots that learn. A robot whose 'brain' was merely a library of predetermined behavior routines, each imagined in full detail by the programmer, would indeed be uninteresting. But such a robot would be incapable of learning anything that the programmer did not know, and thus would fail to be psychologically isomorphic to the programmer, or to any human. On the other hand, if the programmer constructs a robot so that it will be a model of certain psychological laws, he will not, in general, know how it will behave in real-life situations, just as a psychologist might know all the laws of human psychology, but still be no better (or little better) than any one else at predicting how humans will behave in real-life situations. Imagine that a robot at 'birth' is as helpless as a new born babe, and that it acquires our culture by being brought up with humans. When it reaches the stage of inventing a joke, and we laugh, it is simply not true that we are 'appreciating the wit of the programmer.' What the programmer invented was not a joke, but a system which could one day produce new jokes" (H. Putnam, 1964, p. 679).
References
Hilary Putnam. 1964. The feelings of robots. Analysis 19(3), January 1959: 64-68.