Day 1: Thu, June 22
Economic impacts/ Session chair: Jacob Steinhardt
09:00-09:15 Jacob Steinhardt Opening Remarks
09:15-10:00 Michael Webb AI and Automation
10:05-10:50 Dhruv Madeka The Value of Human Level AI to Education
Coffee Break
11:05-11:50 Tom Davidson Will AI lead to transformative economic growth?
12:00-3:15 Lunch/Break
Alignment / Session chair: Amanda Askell
03:15-04:00 Adam Kalai The Consequences of High-Quality Human Simulators for Alignment
Coffee Break
04:15-05:00 Ellie Pavlick Unpacking the “Human” Part of “Human-Level AI”
05:05-05:50 David McAllester Interpretability, Control, and Retrieval Models
06:00-08:00 Dinner
Day 2: Fri, June 23
Security and Policy Implications / Session chair: Michael Webb
09:15-10:00 Nicholas Carlini How could HLAI be used to compromise software systems?
10:05-10:50 Helen Toner Policy approaches to managing risks & harms of AI (including HLAI)
Coffee Break
11:05-11:50 Ajeya Cotra Will HLAI take off “fast” or “slow”?
12:00-01:00 Lunch
01:00-02:25 Break
Imagining Future Systems / Session chair: Irina Rish
02:25-03:10 Jascha Sohl-Dickstein More intelligent agents behave less coherently: the "hot mess" theory of AI misalignment
03:15-04:00 Roger Grosse LLMs, Compression, and Generalization
Coffee Break
Catastrophic Risks / Session chair: Amanda Askell
04:15-05:00 Richard Ngo TBA
05:05-05:50 Dan Hendrycks An Overview of Catastrophic AI Risks (paper: An Overview of Catastrophic AI Risks)
06:00-08:00 Buffet Dinner
Day 3: Sat, June 24
Forecasting / Session chair : David Duvenaud
09:15-10:00 Tamay Berisoglu Projecting future computational resources, data, and algorithms for ML
10:05-10:50 Jacob Steinhardt What will GPT-2030 Look Like?
Coffee Break
11:05-11:50 Jonathan Mann Forecasting Existential Risk
12:00-01:00 Lunch
01:00-01:25 Break
Human-Compatibility / Session chair: Sham Kakade
01:25-02:10 Fernanda Viegas Exploring AI Dashboard Design: The System Model and the User Model
Coffee Break
02:25-03:10 Irina Rish On Scaling Laws, Emergent Behavior, and AI Democratization
03:15-04:00 Dean Foster Using the Principal / Agent problem from Game Theory to think about Alignment
04:10-04:40 Lightning Talks: David Krueger Safe... and Trustworthy... and Responsible AI
Jacob Andreas High-Recall Interpretability
David Bau How can we avoid a decades-long research delay in interpretable ML?
Adam Kalai Intelligence Dimmer
Jacob Steinhardt The Future is (Very) Uncertain
Yejin Choi An Experimental Thought: Aligning AI Through Human Norms, Morals and Values
Stella Biderman Transformer Math
Jascha Sohl-Dickstein Adversarial examples influence both human and machine perception
06:00-08:00 Buffet Dinner