As the use of artificial intelligence becomes more prevalent in our society, and we are increasingly dependent on tech companies to handle our personal data and make the best choices in good faith, it’s evident there is still a lot of progress to be made with how we collectively agree on what is “ethical”, and how we enforce those principles.

I found the paper from the University of Cambridge, “The Role and Limits of Principles in AI Ethics” by Jess Whittlestone et al., compelling because it talked about how there is a general consensus as to what some of the core issues and values of AI ethics are, but the authors argue that agreeing on these generalized values is not enough, and there is still a lot of nuance in how they are implemented.

While there’s agreement in the five principles used in bioethics for AI: “autonomy, beneficence, non-maleficence, justice, and explicability”, these principles are likely to conflict with one another in practice and create “tensions”, where “two values or goals conflict and it is not possible to get more of one without sacrificing the other” (Whittlestone et al. 2019). The authors mention using AI to diagnose medical conditions with more accuracy than a doctor, “for which a ‘full and satisfactory’ explanation cannot necessarily be provided”, but imagine receiving a late stage cancer diagnosis, only to find out later the results of the test were a false positive that was not caught because people’s inability to tune and explain the model properly.

Having ethical principles in place that are widely agreed on is a baseline starting point, but it cannot be left to each individual’s interpretation due to it’s generality, differences in interpretation, and the tensions that arise between the ethics principles in practice. This is made worse by the dangers of allowing the corporations who employ these algorithmic technologies to also be the arbiters of what is ethical and what is not, as Maciej Ceglowski points out in “The New Wilderness”. Ceglowski states it succinctly when he says “their core business activities are not compatible with their professed values. No amount of eloquence can reconcile the things they say with the things their companies do”. It is for these reasons, that I believe we should impose stricter regulations on what is “ethical” for the use of artificial intelligence, similar to the Wilderness Act of 1964, mentioned by Ceglowski, which established protections for certain public lands across the United States to be free of human activity. This current age of artificial intelligence calls for AI ethics regulations free from those who are creating the algorithms.