To make deliberate progress towards more intelligent and more human-like
artificial systems, we need to be following an appropriate feedback signal: we
need to be able to define and evaluate intelligence in a way that enables
comparisons between two systems, as well as comparisons with