Over the last decade the discourse on AI regulation as increased significantly driven by a deeper understanding on the potential impacts that may surface due to the application of AI. Globally, countries have taken different path towards the regulation of AI. Is regulation of AI needed? What are the technical implementation and practical considerations? In this session we will review the how do we bring the implementation of AI regulation to the ‘real-world’.
Recommended Readings and References:
https://artificialintelligenceact.eu/
https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence
https://www.wired.com/story/china-regulate-ai-world-watching/
Automated content moderation, often seen as the solution to problems of scale, inherently requires both policy and technical decisions that have far-reaching impacts on the users on a platform. These include, for example, decisions about how to operationalize content violations (e.g., what constitutes misinformation) and how to calibrate thresholds for flagging content, which result in more false negatives or false positives. Moreover, given the inevitability of imperfect algorithms, how to best implement policy around them also presents a critical problem to solve.
Regulatory proposals for AI oversight are often rooted in Fairness, Accountability, and Transparency: Many proposals aim to prevent discrimination and preserve equity by holding companies accountable through transparency or documentation requirements. What are the practical limits of some of these proposals and what are the friction points where companies and policymakers may need to find a compromise?
NIST is focused on developing consensus technical guidance for managing AI risks that can be applied across contexts regardless of industry. Achieving this goal requires transforming the way we think about and approach the development and deployment of AI systems. This will entail new strategies for broadening perspectives beyond the current computational focus to more cultural and societal factors, and for how we can all get better at producing and evaluating scientific claims.
Optional resources/materials: