Recordings during the AFFINE Superintelligence Alignment Seminar 2026, posted on YouTube. Slides are linked where the speaker shared them.
Recordings · Spring 2026
Talks from the first seminar
Week 1
-
Writing: Getting Started
-
Effective Disagreements 2: Acknowledging Biography
-
Effective Disagreements 1: Socratic Questioning
-
Variants of the Alignment Problem
Week 2
-
Agent Foundations: How and Why
-
Inspection of Thinking as a Source of Ideas for Understanding Minds
-
Mental Mislabelling and How to Fix It
-
The Many Faces of Goodhart’s Law
-
Understanding Understanding (including Peer Tuition)
-
The Trouble with Maximizers
-
Jonas’ Research Agenda
-
Superintelligence Alignment: Principles, Attempts, Contributions
-
Panel: How and Why to Not Burn Out
-
Decision Theory for AI Safety
-
Intro to Infra-Bayesianism
-
Solomonoff Induction and Inductive Biases of Neural Networks
Week 3
-
AFFINE Tech Tree
-
Verification-Based Alignment Schemes
-
Panel: The Point of AFFINE
-
Overview of the Learning-Theoretic Agenda
-
What Is the Alignment Problem?
-
Deliberate Research Practice: Research Insights from Building Superhuman Poker AI
-
Catastrophic Goodhart as a Blind Spot for Computer Science
Week 4
-
Towards a Theory of Identity
-
The Lethal Reality Hypothesis
-
Biology and Habits of Open-Ended Research
-
Safety by Design via Formally Enforcing Care for Human Empowerment?
-
On Model-Wise Thinking
-
Reformulating Strength and Generality of Instrumental Convergence
-
Value Inversion / Pessimization
-
Flaws of Expected Utility Theory and Ways to Avoid Them
-
QACI: Question-Answer Counterfactual Interval