Publication | Conference on Neural Information Processing Systems 2022

Communicating Natural Programs to Humans and Machines

Four ARC tasks, the goal is to correctly infer the unseen output from the given examples.

This study was conducted to understand how humans use language to instruct each other in order to perform specific tasks. Autodesk researchers found that many tasks can be instructed using language (e.g., “Can you align all the bathroom stalls on this floor?”), while the output must be very specific.

This paper was presented at the Conference on Neural Information Processing Systems 2022.

Download publication

Abstract

Communicating Natural Programs to Humans and Machines

Samuel Acquaviva, Yewen Pu, Marta Kryven, Theodoros Sechopoulos, Catherine Wong, Gabrielle E Ecanow, Maxwell Nye, Michael Henry Tessler, Joshua B. Tenenbaum

Conference on Neural Information Processing Systems 2022 (Featured Presentation)

The Abstraction and Reasoning Corpus (ARC) is a set of procedural tasks that tests an agent’s ability to flexibly solve novel problems. While most ARC tasks are easy for humans, they are challenging for state-of-the-art AI. What makes building intelligent systems that can generalize to novel situations such as ARC difficult? We posit that the answer might be found by studying the difference of language: While humans readily generate and interpret instructions in a general language, computer systems are shackled to a narrow domain-specific language that they can precisely execute. We present LARC, the Language-complete ARC: a collection of natural language descriptions by a group of human participants who instruct each other on how to solve ARC tasks using language alone, which contains successful instructions for 88% of the ARC tasks. We analyze the collected instructions as ‘natural programs,’ finding that while they resemble computer programs, they are distinct in two ways: First, they contain a wide range of primitives; Second, they frequently leverage communicative strategies beyond directly executable codes. We demonstrate that these two distinctions prevent current program synthesis techniques from leveraging LARC to its full potential and give concrete suggestions on how to build the next-generation program synthesizers.

Associated Researchers

Samuel Acquaviva

MIT

Marta Kryven

MIT

Theodoros Sechopoulos

MIT

Catherine Wong

MIT

Gabrielle E Ecanow

MIT

Maxwell Nye

MIT

Michael Henry Tessler

MIT

Joshua B. Tenenbaum

MIT

View all researchers

Get in touch

Something pique your interest? Get in touch if you’d like to learn more about Autodesk Research, our projects, people, and potential collaboration opportunities.

Contact us