A Domain-Specific PPL for Reasoning about Reasoning - A Memo on Memo

This 20-minute video presents a conference talk from the LAFI workshop (January 19, 2025) where researchers from MIT introduce "memo," a domain-specific probabilistic programming language designed specifically for theory of mind applications. Explore how this innovative language addresses the challenges of modeling recursive probabilistic reasoning about thinking processes. The presentation demonstrates how memo provides specialized syntax and semantics that make it easier for programmers to express complex theory of mind concepts while leveraging array programming for significantly improved performance on modern hardware. Learn how this approach has enabled researchers to create faster models with less code, leading to its adoption by multiple research groups. The talk, delivered by Kartik Chandra, Tony Chen, Joshua B. Tenenbaum, and Jonathan Ragan-Kelley, offers valuable insights for those interested in computational modeling of human cognitive processes.