The academic year in Princeton is starting, so I have enrolled to audit the computer science course COS 324 – Introduction to Machine Learning. I have no intention to become the next bright spark in artificial intelligence. I’ve joined this course, though, because I feel that many of the legal, policy, and philosophical analyses we read about machine learning are poorly informed. While it may be true that one doesn’t need to be able to operate a tractor to make sensible agricultural policy, I do not think it’s healthy to regulate complex algorithms without understanding the potential and limitations of these algorithms. Ian Bogost argued, rightly in my opinion, that the blind fascination with algorithms has reached a near theological level, which may better be brought back down to Earth. Only then will lawyers, policy makers, and philosophers be able to make meaningful sense of them.
Impact of AI engineering
The knowledge gained through this course will also be relevant for our new project. We aim to understand how decisions in machine learning engineering can affect the social model we’re building towards. In other words, we aim to link the engineering ethics and political theory impacts of AI engineering. Understanding this link will allow us to develop the policies that incentivize engineers, companies, and governments to optimize their machine learning algorithms not just for efficiency and speed, but also concepts such as justice and fairness, applied to their specific contexts of deployment. Our first publications in January of next year will showcase some of this thinking in concrete cases.