Ethical Problems of Contemporary AI

The logic-based AI which we considered in Part II had many notable successes. It produced the first AI world chess champion, discovered the proof of a mathematical theorem which eminent human mathematicians had tried in vain to prove, discovered some new laws of nature in the natural sciences, and made progress with automating medical diagnosis. Despite these successes, logic-based AI programs did not have many applications in industry and commerce. There were a few such applications. In section 5.2 , we mentioned an example from Quinlan in his 1986 paper (p. 85). This was the use of a logic-based AI program (a derivative of ID3) by Westinghouse Electric’s Water Reactor Division in a fuel-enrichment application, which boosted the company’s revenue by more than ten million dollars per annum. However, these few applications of logic-based AI were relatively unproblematic. The situation changed dramatically with the beginning of the deep learning revolution in 2012. The novel neural-network-based AI turned out to have a whole range of applications in industry and commerce. Unfortunately, many of these applications are widely considered to be morally wrong. Consequently, the philosophy of AI now includes ethics.