I would like to warmly compliment Prof. Choucri for a masterly compilation of a concise, and yet comprehensive, paper outlining the elements of an AIIA, capturing the intricacies of the issues involved and appreciating the breadth of the challenges.
She makes the absolutely valid point that transformation of AI innovations and applications are moving ahead of our collective ability to fully appreciate the opportunities and threats that they generate. It is an equally valid point that, while countries are developing their own AI policies (at different levels of sophistication), there has been very little realization of the urgent need for an international regime that promotes cooperation on AI applications for human progress and mitigating their destructive potential.
I am almost entirely in agreement with the deliverables she has identified and the methodology of approach to them. What I will add here are comments on some of the aspects covered in the draft.
The need for inclusiveness in this international accord merits particular emphasis. The borderless nature of AI tools means that those excluded would have no incentive not to use them to the detriment of others. The need to maximize the number inside the tent is obvious.
It follows from this that we should not limit AIIA to democracies. In any case, there are too many “models” of democracy these days and variances even within each. The important criterion should be the willingness to abide by agreed norms; not the system of governance. Vietnam, for example, is not called a democracy, but is a disciplined member of the international community, abiding by international law. Exclusion of Vietnam and countries like it, on the grounds of their domestic system of governance, would be counterproductive.
The draft rightly assigns to national authorities the discretion to determine the manner of AI governance within their countries, while requiring them to abide by the discipline of the international accord. This is wise. As we see today, even among democracies, there are different practices and different interpretations of the ethics of reconciling the conflicting interests of human rights, law enforcement and preventive security measures. The balance that each country will find would depend on its historical experiences, social consciousness and threat perceptions. Prescriptive provisions threaten disruption of the international accord.
A cautionary note on sanctions that are envisaged to enforce adherence to the Accord. Recent experience of unilateral primary and secondary sanctions underlines the need for carefully defining the scope and applicability of sanctions, identifying clearly the objective conditions which would trigger the sanctions, and the methodology of applying and lifting the sanctions.
The draft makes the important point of variances in the levels of AI-capability across the world and the need for capacity-building. We need to be conscious of the dangers of a new digital divide, which would widen global social and economic inequalities, with consequent threats to peace and security.