Can AI Take Part in Dialogue?

Abstract

Many properties, thought to be proper to human intelligence only, have been programed. Some are still considered by many to be resistant to simulation – for instance recognizing the truth of Gödel’s sentence; this is, however, unjustified. Other general human characteristics can be hard to simulate, for example creativity. Yet we cannot rule out the possibility that each human quality can be simulated to be a feature of robots. To be sure, doubts remain: simulation is not enough, many of us would say. Perhaps, but one can never be sure since it can be good enough for all tangible measurements. There exists, however, a possibility of showing human superiority that seems more hopeful. Namely, it would be easier to point out human specificity when we enter the interhuman level. Then relations are considered rather than qualities of individual humans. An ultimate strengthening of the Turing test can illustrate this idea; it is about raising children. Another, more realistically accessible general example of the relational approach is dialogue. Genuine dialogue could turn out to be inaccessible to robots because simulated dialogue is not real. The reason is that for the robot everything is an object, while according to the philosophy of dialogue, humans operate in a qualitatively different “I-Thou” dimension. And, not less important, unlike humans, two robots seem to form essentially one larger robot.