Research & talks
Can we read an AI’s mind?? "Dilly-dallying...", "Moonwalking..." and more
AI safety researcher Julian Schulz will unpack mechanistic interpretability, steganographic reasoning, and whether we can trust what LLM chains of thought reveal, followed by an open Q&A. A MATS scholar and Visiting Researcher at Meridian, he applies interpretability at scale and studies how models may learn hidden reasoning.
Explore event