Robots that resemble human beings can have practical applications (humanoid robots) or they can be a new science of human beings that will allow science to understand human beings as it understands everything else (human robots). While humanoid robot physically resemble human beings and do only a few things that human beings do, human robots must do everything that human beings do and, since human beings have free will, human robots must have free will, and my talk will describe the first steps towards the construction of robots that have free will. We must begin by recognizing that the most important difference between humanoid robots and human robots is that, while humanoid robots must do what we want them to do, human robots must do what they want to do. What is the difference between doing X and doing X because one wants to do X? Robots that have free will are robots with an artificial brain which is able to predict the consequences of their actions before actually executing the actions, to judge these consequences as good or bad (where judging something as good or bad is a spontaneous result of biological evolution and learning, including learning from others), and to actually executing an action only if its consequences are judged as good. In addition, robots that have free will must possess a human-like language with which they can not only talk with other robots but also talk with themselves, and they must use this language to articulate and reason about the consequences of their actions. Living with robots that have free will, either physically realized or even only simulated in a computer, will pose more serious problems to human beings than living with today’s robots which do only what they have been programmed to do. In fact, in the future human beings will be confronted with a very difficult choice: either accepting to deal with these problems or renouncing to understand themselves as science understand everything else.
