Skip to Main Content
Lightning Talk Beginner MIT

LeanDetective - Can AI solve murder mysteries?

Proposal status is Rejected
Session Description

I will talk about LeanDetective, an open-source Lean4 library I created which acts as a Domain-specific language to model detective mysteries such as Drishyam or Knives Out.

Lean4 uses Dependent Type Theory, where propositions become types, and programs become proofs. This makes it possible to prove properties of programs within the same environment.

I will introduce basic techniques like inductive types, modeling spatio-temporal histories, pluggable physics models and using them to establish theorems like "Prime suspect is whoever had the motive, means and opportunity".

This also gives a glimpse on how math and programming languages will increasingly open more domains and roles to AI agents, such as crime detectives, lawyers and perhaps judges.

Key Takeaways
  • Intro to Lean4 & Dependent Types

  • How to model murder mysteries in LeanDetective

  • Showcase examples: Murdle, Drishyam & Knives Out

References

Session Categories

Introducing a FOSS project or a new version of a popular project
Technology architecture
Talk License: MIT

Which track are you applying for?

Compilers, Programming Languages and Systems

Speakers

Nilesh Trivedi
Nilesh Trivedi Co-founder and CTO
  • 20+ years in tech

  • Co-founder and CTO for a startup in AI

  • Creator of Fixable wiki, SWE-WebDevBench, SwadeshiAlternative and other FOSS projects.