--- title: 'Do As I Can, Not As I Say: Grounding Language in Robotic Affordances' license: CC-BY-4.0 license_url: https://creativecommons.org/licenses/by/4.0/ license_note: 'This note is CC BY 4.0. It links to the paper and does not reproduce it. Paper license: arXiv non-exclusive distribution (verified on arXiv abs page 2026-11-11; http://arxiv.org/licenses/nonexclusive-distrib/1.0/).' source: https://iinventedlight.com/library/notes/2204-01691-saycan.md source_license: 'arXiv default (non-exclusive distribution license)' attribution: 'Pointer note by iinventedlight (AI-assisted). Links to: Michael Ahn, Anthony Brohan, Noah Brown, Yevgen Chebotar, Omar Cortes, et al. (45 authors), Do As I Can, Not As I Say: Grounding Language in Robotic Affordances (arXiv). This note does not reproduce the paper.' author: 'iinventedlight (AI-assisted)' date: 2026-11-11 status: published approved_by: "Cole Withington, 2026-11-11 20:58 PT, Grok Bot chat widget t1717s7407: \"Publish all 100 including the LLMs Encode Clinical Knowledge note\"" checked_on: 2026-11-11 as_of: 2026-11-11 confidence: medium untested: 'Pointer note only. Abs page read 2026-11-11. The paper is not reproduced here.' category: arxiv topic_area: 'ai-research: tools' tags: arxiv, pointer, tools, robots arxiv_id: 2204.01691 alias: 2204.01691-saycan format: pointer type: pointer sources: - id: S1 title: 'Do As I Can, Not As I Say: Grounding Language in Robotic Affordances' publisher: arXiv authors: 'Michael Ahn, Anthony Brohan, Noah Brown, Yevgen Chebotar, Omar Cortes, et al. (45 authors)' url: https://arxiv.org/abs/2204.01691 version_read: 'v2 (16 Aug 2022)' accessed: 2026-11-11 license: 'arXiv default (non-exclusive distribution license)' --- # Do As I Can, Not As I Say: Grounding Language in Robotic Affordances ## Summary Do As I Can, Not As I Say: Grounding Language in Robotic Affordances (2022), by Michael Ahn, Anthony Brohan, Noah Brown, and many coauthors, is at https://arxiv.org/abs/2204.01691. Their method uses pretrained skills to constrain the model to propose natural language actions that are both feasible and contextually appropriate. In this setup, the robot supplies a form of hands and eyes, while the language model supplies high-level knowledge about the procedure, and value functions attached to the skills provide grounding in the physical environment. They evaluate the approach on real-world tasks with a mobile manipulator. The authors say a significant weakness of language models is that they lack real-world experience, which makes it difficult to leverage them for decision making within a given embodiment. They evaluate the method on a number of real-world robotic tasks, and they report that this approach is capable of completing long-horizon, abstract, natural language instructions on a mobile manipulator. The project website named in the abstract is https://say-can.github.io/. Title, authors, abstract, and comments were taken from the arXiv abs page. The version is v2 (16 Aug 2022), submitted 4 Apr 2022. ## Claims Pointer notes do not quote the paper. See Verification and Sources for the link and license. ## Verification - Format: pointer note only. The paper is not reproduced here. - Why it matters: It matters as a key paper connecting language models to physical action by constraining plans to what a particular robot can actually do. - Reproduced: not reproduced (2026-11-11). - Related work: Related notes already in the library include [ReAct](https://iinventedlight.com/library/notes/2210-03629-react-reasoning-and-acting/) and [CodeAct](https://iinventedlight.com/library/notes/2402-01030-codeact/). - Limits: This is a pointer note only. It summarizes the grounding idea at a high level and does not describe robot control procedures. The project site is named in the paper comments. Project site named in the comments: https://say-can.github.io/. - License: arXiv default (non-exclusive distribution license) on the abs page (2026-11-11; http://arxiv.org/licenses/nonexclusive-distrib/1.0/). Submitted 4 Apr 2022 (v1); last revised 16 Aug 2022 (v2). Sources: [S1] https://arxiv.org/abs/2204.01691, v2 (16 Aug 2022). checked_on: 2026-11-11. Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it. The note itself is CC BY 4.0. ## When to apply Use when you need the original paper on grounding language-model plans in a robot's learned skills and affordances. ## Facts - id: F1 claim: 'The paper proposes grounding language models with pretrained skills so that proposed natural language actions are feasible and contextually appropriate for a robot.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0 - id: F2 claim: 'Value functions associated with those skills are used to connect high-level language-model knowledge to a particular physical environment.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0 - id: F3 claim: 'The authors evaluate the method on real-world robotic tasks with a mobile manipulator, and they point to https://say-can.github.io/.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0 ## Sources - [S1] arXiv 2204.01691, https://arxiv.org/abs/2204.01691. Paper license: arXiv default (non-exclusive distribution license). v2 (16 Aug 2022). Accessed 2026-11-11. This note links to the paper and does not reproduce it. License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it. --- Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).