This contribution synthesizes the discussions of the special session on policies for Artificial Intelligence in Science and Innovation, organized by the OECD’s Directorate for Science, Technology and Innovation. The session was opened by Dr Judith Arrieta, Minister of the Foreign Service at the Chief of Staff’s Office of the Secretary of Foreign Affairs of Mexico, and the two panels included speakers from governments, industry and civil society from European countries, USA,Canada China and Australia. Participants discussed the disruptive nature of AI and the formidable challenges it poses. Most of the discussion focused under the umbrella title of ethics, but they span very different issues of human-centered values, fairness, transparency, explainability, and many more. Other challenges include employment, education, SME policy, enabling environment, access to data and computing technology. Responses by governments were also discussed with a particular focus on national strategies, whose main pillars are oriented toward knowledge creation through AI research, knowledge diffusion through linkages to the private sector, development of human capital which will underpin the development of the sector, and a strong values, ethical and regulatory framework to create the conditions for the development of trustworthy AI.
In a world of finite resources, discussants concluded that one cannot apply very stringent requirements to all AI decisions, and there is clearly a need to require more transparency, explainability and robustness from systems which have the greatest impact on human lives. Therefore an approach based on algorithmic impact assessment seems reasonable. Such an approach needs to be further developed and standardized.