A machine that knows its own code (Q456969)

A philosophical perspective on artificial intelligence naturally leads to the question whether it is possible to know that one is a certain machine. As known statements should be true, the interesting question is then whether a machine can know that it is a certain machine. With a suitable formalization of what it means for a machine to know something (which is possible, e.g., in Shapiro's `epistemic arithmetic'), this can be turned into a precise mathematical question that has been studied by various authors and answered in the negative. Roughly, one introduces a modal operator \(K\) for `knowing' and then asks whether \(K(\phi)\) can be true for some \(e\), where \(\phi\) is the statement that \(K(\psi)\) holds for a certain statement \(\psi\) iff \(\psi\) is contained in the \(e\)th recursively enumerable set. An assumption common to the considerations so far is that the machine knows that what it knows is true. The paper contributes to this area of study by dropping this assumption. It turns out that in this setting, a machine can actually `know' its own code. The paper is well written and mostly self-contained; only a basic knowledge of logic and recursion theory is required on the part of the reader.