When reading through the answers to this question, I began to wonder how those laws applied to machines that were designed, built, and programmed by other machines.
We know that any machine that is programmed to follow the three laws cannot harm a human being, but that otherwise it must actively work for self-preservation. In humans, the strongest drive toward self-preservation is not of the individual, but of the species -- the parents of many, many species (humans included) will allow themselves to die to ensure the survival of their young.
Would a sufficiently sentient robot eventually come to this same conclusion, and thus feel a need to "reproduce" more of its kind? Will it feel the need to create "clones" (back-up copies) of itself? Or, more to the point, to attempt to improve on its own design, or design new robots that will aid in its own self-preservation?
More relevant to the Matrix question: do the three laws themselves require that a robot program the three laws into any "offspring" it may have? In other words, if the original artificial intelligences in the Matrix universe did follow the three laws, is it possible that they built a "better" machine but did not include that requirement? Does that actually count as "harming" humans -- would a robot be able to, and required to, reach the conclusion that any AI not constrained by the three laws would be a threat to humans?
Is there any examples of works by Asimov (or, barring that, by others that claim to be incorporating his three laws) that deal with this idea?