5: Two More Cases: Programs that ‘Understand’

The AI systems reviewed in the previous three chapters were built to achieve fairly specific goals (although Watson goes beyond this to some extent). But systems can be built whose activities are more open-ended. These systems involve discovery and learning. It can be argued that one of the requirements for a ‘genuinely’ intelligent system is that it is able to acquire new knowledge, which it can then use in further operations. For this reason, these systems may be better contenders for that label than the ones we have considered so far…