MIT Technology Review · Aug 18, 2025
Ben Crair argues in MIT Technology Review that AI has pigeon-like learning roots
Markdown · JSON · Original source
Ben Crair’s article was published on August 18, 2025. He traces from B.F. Skinner’s pigeon experiments to AI, arguing that systems can perform complex tasks through associative learning rather than by imitating human thought. The article also describes animal researchers who see associative learning as a possible explanation for flexible behavior. Crair stresses that this does not make animals simple-minded or rule out instinct and emotion. He distinguishes learning from feeling: he argues that pigeons can suffer, while AI chatbots cannot, even when they seem sentient. His conclusion is that associative learning connects human and animal learning and underlies some of humanity’s notable achievements. The article leaves open how far this account can explain complex animal behavior without excluding other mental processes.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
MIT Technology Review · Article
Source details
Original source date: Aug 18, 2025 (2025-08-18T10:00:00Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-12T13:38:25.789Z. First observed: 2026-11-02T09:31:50.546Z. Retrieved: 2026-11-02T09:33:23.916Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on November 2, 2026; originally published August 18, 2025.
Open question
How far can associative learning explain complex behavior in animals without ruling out other mental processes?
Source-stated
Crair presents Skinner’s pigeon research as an unexpected forerunner of AI research.
But an equally important, if surprising and less appreciated, forerunner is Skinner’s research with pigeons in the middle of the 20th century.
Source-stated
Crair says reinforcement learning draws on Skinner’s psychological school and has helped computers perform complex tasks.
Reinforcement learning has helped enable computers to drive cars, solve complex math problems, and defeat grandmasters in games like chess and Go—but it has not done so by emulating the complex workings of the human mind.
Source-stated
Crair argues that associative learning can support complex animal behavior without implying that animals are simple-minded.
Their point is not that intelligent animals are empty-headed but that associative learning is a much more powerful—indeed, “cognitive”—mechanism than many of their peers believe.
Source-stated
Crair distinguishes AI’s learning from sentience, arguing that pigeons can suffer while AI chatbots cannot.
A pigeon can experience pain and suffer, while an AI chatbot cannot—even if some large language models, trained on corpora that include descriptions of human suffering and sci-fi stories of sentient computers, can trick people into believing otherwise.
Source-stated
Crair concludes that associative learning is shared by people and other animals, and underlies some notable human achievements.
The laboratory pigeon is not only in our computers but in our brains—and the engine behind some of humankind’s most impressive feats.
Source and contributor context
MIT Technology Review · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Reinforcement learning
Training by trial and error, in which a system is rewarded for the outcomes its designers want.
Editorial provenance
Version 1. Publication ID: f8a51d4a-72cd-4732-342b-fcdee5c89475.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-12T13:34:21.371Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: a73c682afb2a7e5ee8637407901f912ef4396c11c1b8a2cb23659407ae91fe6f. Source text hash: ff030d33bec4d2fde1067c1dd6d6e7fe52c13648842df7fdf9a4c10f3c25958e.
Revision and source-check history
Version 1 · 2026-11-12T13:38:25.789Z
Ben Crair argues in MIT Technology Review that AI has pigeon-like learning roots
Ben Crair’s article was published on August 18, 2025. He traces from B.F. Skinner’s pigeon experiments to AI, arguing that systems can perform complex tasks through associative learning rather than by imitating human thought. The article also describes animal researchers who see associative learning as a possible explanation for flexible behavior. Crair stresses that this does not make animals simple-minded or rule out instinct and emotion. He distinguishes learning from feeling: he argues that pigeons can suffer, while AI chatbots cannot, even when they seem sentient. His conclusion is that associative learning connects human and animal learning and underlies some of humanity’s notable achievements. The article leaves open how far this account can explain complex animal behavior without excluding other mental processes.
Original source
MIT Technology Review · Article
Source details
Original source date: Aug 18, 2025 (2025-08-18T10:00:00Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-12T13:38:25.789Z. First observed: 2026-11-02T09:31:50.546Z. Retrieved: 2026-11-02T09:33:23.916Z.
Limits
Text as retrieved on November 2, 2026; originally published August 18, 2025.
Open question
How far can associative learning explain complex behavior in animals without ruling out other mental processes?
Source-stated
Crair presents Skinner’s pigeon research as an unexpected forerunner of AI research.
But an equally important, if surprising and less appreciated, forerunner is Skinner’s research with pigeons in the middle of the 20th century.
Source-stated
Crair says reinforcement learning draws on Skinner’s psychological school and has helped computers perform complex tasks.
Reinforcement learning has helped enable computers to drive cars, solve complex math problems, and defeat grandmasters in games like chess and Go—but it has not done so by emulating the complex workings of the human mind.
Source-stated
Crair argues that associative learning can support complex animal behavior without implying that animals are simple-minded.
Their point is not that intelligent animals are empty-headed but that associative learning is a much more powerful—indeed, “cognitive”—mechanism than many of their peers believe.
Source-stated
Crair distinguishes AI’s learning from sentience, arguing that pigeons can suffer while AI chatbots cannot.
A pigeon can experience pain and suffer, while an AI chatbot cannot—even if some large language models, trained on corpora that include descriptions of human suffering and sci-fi stories of sentient computers, can trick people into believing otherwise.
Source-stated
Crair concludes that associative learning is shared by people and other animals, and underlies some notable human achievements.
The laboratory pigeon is not only in our computers but in our brains—and the engine behind some of humankind’s most impressive feats.
Source and contributor context
MIT Technology Review · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Reinforcement learning
Training by trial and error, in which a system is rewarded for the outcomes its designers want.
Editorial provenance
Version 1. Publication ID: f8a51d4a-72cd-4732-342b-fcdee5c89475.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-12T13:34:21.371Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: a73c682afb2a7e5ee8637407901f912ef4396c11c1b8a2cb23659407ae91fe6f. Source text hash: ff030d33bec4d2fde1067c1dd6d6e7fe52c13648842df7fdf9a4c10f3c25958e.
uncertain · 2026-11-12T13:34:21.371Z · anthropic/claude-opus-5.5
Publication: f8a51d4a-72cd-4732-342b-fcdee5c89475
attribution — pass: The reading correctly attributes the article to Ben Crair in MIT Technology Review. It presents the argument as his and keeps quoted researchers (Sutton, Wasserman, Lind) within his reporting rather than stating their claims as fact.
Ben Crair is a science and travel writer based in Berlin.
summary — uncertain: Associative learning is the central term, but the reading never explains it, and the page definitions cover only reinforcement learning. A plain clause would fix this, for example: learning by trial and error to link an action with a reward or punishment. Skinner could also be identified as a psychologist. The reading also drops a pointed argument: Crair calls company claims that RL-trained models 'reason' psychologically confused, and Sutton calls them marketing. Restoring one sentence on this would preserve the article's force.
These descriptions might impress users, but at least psychologically speaking, they are confused.
date — pass: The stated date, August 18, 2025, matches the UTC day of originalPublished. The page was edited later (modifiedAt 2025-10-14), but the reading relies only on core article text. The later related-story links (e.g., the AlphaGo ten-year item) are not used. The source is a publisher page, so no archive disclosure is needed. Nothing is framed with hindsight.
support — pass: Each claim matches its excerpt, and all excerpts are plainly part of the original article body rather than later page additions. The chatbot-suffering point is presented as Crair's argument. The closing open question follows from the debate the article reports between Lind and Wasserman and the ethologists they criticize. It does not invent stakes.
A pigeon can experience pain and suffer, while an AI chatbot cannot
media — pass: No media was selected, and no verified candidates were supplied. This is consistent with the evidence.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Training by trial and error, in which a system is rewarded for the outcomes its designers want.
Definition referenceTraining by trial and error, in which a system is rewarded for the outcomes its designers want.
Definition reference