I so love the UNESCO AI "ethics" recommendation. It's here: https://www.unesco.org/en/artificial-intelligence/recommendation-ethics It's really way more than an AI document. It's more like a general pathway forward on peaceful, sustainable development. AI is the new paper, on which we write out our intentions and procedures, incorporating and expressing our values more explicitly than we ever have had to before. Before, our laws were always intermediated by people, so we could make assumptions (without even noticing) about "common sense," accidentally as we did so propagating biases and so forth. Neither AI nor anything else can entirely get us away from that, but by making our rules more explicit so that they can be at least in part executed by machines, we are putting ourselves in the place where we can better deal with the sources of bias in our culture.
|Also contains great developer- & |
user-focussed images for AI!
Based on my quick reading of early drafts of the recommendation, formerly thought the document's main contribution was really being very clear on how to generate fairness and inclusion in the context of AI. Basically, I saw it as extending and humanising the content academics and even the IEEE had already been developing.
But now that I read it more carefully from the perspective of a professor in a (German!) Policy & Governance university, I see even more. At Hertie School I'm learning that and why (and for good reasons) even the notions of human rights and dignity are being contested as excluding and even colonial. For example, there's an argument that presently 'developed' countries became rich by brutally exploiting their own and other countries' citizens and residents, and now are trying to use human rights to lock in their hold on power and wealth. Among other things, the UNESCO document does incredible work in addressing such concerns, justifying its grounding of the values underlying many ethical statements in the Universal Declaration of Human Rights (also well worth a read.) But the AI Ethics Recommendation goes well beyond the UDHR, really laying not just the foundations but the first few floors for the legal frameworks our nations and states should be developing. I was most familiar with the fact it excluded legal personality for AI (I've been told that this kept me out of the UNESCO room, because my paper on AI legal personality made me somehow be seen as partisan rather than expert on that critical issue.) But all the policy portions, such as those on on transparency, audits, misuse of anthropomorphism, etc. in my opinion lead even the EU's world-leading legislative effort, the draft AI Act (link: PDF of a 2022 paper I wrote about that.)
This document is wholistic; it could almost be the grounding of a new philosophy that extends from the best of the humanities into the constraints for ecologically and socially sustainable economics and government. I'm thinking of restructuring my AI ethics courses to start from reading the UNESCO recommendation and the UDHR, then teach AI & its governance as history leading to this point. Presently though, this document is in my final lectures, where I discuss the various transnational efforts to regulate AI. But rereading it in the context of the book on AI and Ethics I'm writing, I'm rethinking a lot of things. One of the main points I find myself needing to make is that we all constitute not only our nations including our state, but also their government, though obviously depending on the particular state with varying amounts of efficacy. But seriously, everyone who's read this far in this blog, please read also the UNESCO recommendation, then think of all the little ways you can help it get recognised and realised.
|Another great image from the UNESCO AI Ethics Recommendation|