• No-free-will Decision Theory:
  • How do we make decisions when we know that they are fully (or partially?) determined by prior causes (some of which also have other effects than our decision)?
• Con artists and manipulation techniques are adversarial examples for humans.
• Resilience to adversarial examples is in opposition to efficiency (small models, little data, etc.) -- implications for AI safety.
• If we had a perfect crime predictor, what would we do with it? Would we want to use it for prosecution?
  • But we could then change the future!
• Observation: people like the speakers if they understand the talk and can contribute and ask good questions.
• Savage decision theory, Fish...
• Toxin puzzle vs. Parfit's hitchhiker -- same or not?
• Rheinbach common cause principle.
• Women give social justice talks -- seems like gender balance is important?

Closing remarks

• CFI scope: AI impacts including the relevant technical bits.
• CGI mission: create the community of thinkers to look into the shape and impacts of human-level AI, etc.
  • Subproject model: banyan tree metaphor.
• We want to go towards an international community for beneficial AI.

Program

27 July 2018

9:00 Coffee/Tea 9:25 Stephan Hartmann: Opening
9:30 Hans Briegel: Transparency of Classical and Quantum Mechanical Artificial Intelligence 10:45 Tom Sterkenburg: Inductive Bias and Adversarial Data 11:30 Break 11:45 Teresa Scantamburlo: The Progress of Machine Decisions 13:00 Lunch 14:30 Dominik Janzing: Principle of Independence of Mechanisms in Machine Learning and Physics 15:45 Benjamin Eva: The Similarity of Causal Structure 16:30 Break 16:45 Christian List: Representing Moral Judgments: The Reason-Based Approach

19:00 Dinner 28 July 2018

9:00 Coffee/Tea 9:30 Wolfgang Spohn: From Nash to Dependency Equilibria 10:45 Johannes Treutlein: How the Decision Theory of Newcomblike Problems Differs Between Humans and Machines 11:30 Break 11:45 Tina Eliassi-Rad: Just Machine Learning 13:00 Lunch 14:30 Aidan Lyon: Forecasting Forecasting: Humans vs Computers vs Humans + Computers 15:45 Zoe Cremer: Preserving AI Control via the Use of Theoretical Neuroinformatics in Empirical Ethics 16:30 Huw Price: Closing