In our lives we are going to encounter more machines that are more ‘intelligent’. In this chapter I will try to explain what this means and how we should think about it. My starting point is ideas drawn from recent interpretations of Wittgenstein (e.g. Winch, 1958; Bloor, 1983) and the sociology of scientific knowledge (e.g. Collins, 1985). 1 I intend to use these ideas to look at the development of intelligent programs called ‘expert systems’. Expert systems are with us now. They are relatively accessible in the sense that the principle underlying their design is easy to grasp, and in that they are not too difficult to build. I think that the ideas underlying expert systems are essentially the same as the ideas underlying much work in Artificial Intelligence (AI). Even if there are new or different principles of AI that are not encompassed by the model advanced here, I hope to show (implicitly) the kind of difference in design principle that will be required to make an ‘in principle’ in the ability of machines to think like humans. It will not be a difference in quantity.