Several European projects are contributing tools for applied AI ethics. A recent paper in Science and Engineering Ethics from the Human Brain Project suggests that that ethical reflection on the practical aspects of artificial intelligence requires theoretical reflection and clarification of key concepts: What is intelligence? And how do we make the distincition between natural and artificial intelligence? Especially if an artificial intelligence is to replace a human being.
In the paper, Michele Farisco, Kathinka Evers, and Arleen Salles from Uppsala University suggest a methodological model for ethical analysis of instances where AI is overtaking human-specific activities. According to them, stating whether or not AI is “intelligent”, despite a lack of moral features, requires both caution and comprehensive ethical analysis.
To date, state-of-the-art artificial intelligence lacks several ethically relevant and critical features of natural intelligence. For example, an AI has no interests, no preferences, and cannot act autonomously, at least in a morally relevant sense. In many respects, this means that reports of threats posed by the use of AI technology might be exaggerated, or (because at this point in time this intelligence cannot act autonomously) at the very least premature. However, the increasing use of AI in certain domains of society does raise legitimate, ethical questions.
And although there is a great need for an “ethics by design” (where ethical principles are embedded from the very beginning of the design process) of the kind that has been proposed by for example H2020 projects SIENNA and SHERPA, applied ethics will benefit from conceptual clarification. According to Michele Farisco and co-authors, human intelligence has several qualities that AI (still) does not. “Whether this limitation is intrinsic to AI or only a matter of further development, we will find out. Until then, we need to be careful, and avoid opposition to the technology as well as an overly enthusiastic approach to the possibilities of AI”.
Farisco, M., Evers, K. & Salles, A. Towards Establishing Criteria for the Ethical Analysis of Artificial Intelligence. Sci Eng Ethics (2020). https://doi.org/10.1007/s11948-020-00238-w 7 July 2020
By Anna Holm & Josepine Fernow