Towards Steerable AI Systems

Explore the future of AI systems in this keynote address from SIGIR 2024, delivered by Professor Thorsten Joachims. Delve into the critical need for steerable control in AI, from personal assistants to large internet platforms, as these systems increasingly act on our behalf. Examine machine learning methods that enable different stakeholders, including end-users and platforms, to guide AI actions towards beneficial long-term outcomes. Discover models capable of reasoning about goals at multiple scales, balancing short-term and long-term metrics. Learn about practical applications of these methods in search ranking and writing assistance. Gain insights from Joachims, an ACM Fellow, AAAI Fellow, and KDD Innovations Award recipient, as he shares his expertise in counterfactual and causal inference, policy learning, and learning from implicit feedback.