There has been much talk of ethics in AI recently. Some enterprises and vendors and consultants have talked of the need for a chief AI ethics officer. On the face of it, it sounds like a good idea, but I don’t think it is.
Note that I am not arguing against ethical AI or responsible AI
I am saying that at an organization or a vendor, we do not need a single person (or a team) as a gatekeeper for all ethics related to AI. So, if we don’t need a chief AI ethics officer (or a team of AI ethics officers) – what do we need instead?
Let’s consider another area: Privacy
Like ethics, everyone has a view on privacy. Privacy is also a topic that generates a lot of hearing (discussion) but no light (actionable insights). So, what we need is something that is actionable. Also, there are already detailed methodologies for ethics in AI such as from the Alan Turing Institute in the UK and NIST in the USA. So, considering that we have good methodologies – then what we need is ways to implement ethics in a pragmatic sense. Modeled on the idea of Privacy by design, we could consider the idea of Responsible AI by design. Already, some literature exists in this area for Responsible AI by design
But more promisingly, we have toolkits for implementing responsible AI. So, I believe that we don’t need a chief AI ethics officer – rather we need implementations of responsible AI by design at a developer level
Image source: https://pixabay.com/photos/jurisdiction-lady-justice-justice-677940/