Escape Meridian

MIT Technology Review · Nov 27, 2024

MIT Technology Review reports AI agents developed social roles and customs in Minecraft

Markdown · JSON · Original source

MIT Technology Review published this report on November 27, 2024. Altera’s Project Sid placed in Minecraft, where they developed social roles, influenced one another’s votes, and spread memes and a parody religion. The article says these behaviors arose from patterns learned by (LLMs), combined with Altera’s system for turning those patterns into context-aware actions; it also stresses that the agents are not alive. Altera cofounder Andrew Ahn says the results suggest LLMs can mirror human social dynamics. Altera founder Robert Yang sees the experiment as an early step toward agents that could work alongside people, while AI veteran Julian Togelius doubts that such systems experience anything. Togelius nevertheless says a convincing simulation of care might be useful, depending on whether the person receiving care minds that the carer has no experiences. The article’s closing thought is that apparent care might matter even without real feelings. Whether simulated care would have the same value to the person receiving it remains unresolved.

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.

MIT Technology ReviewArticle · technologyreview.comOpen original source

Original source

MIT Technology Review · Article

Source details

Original source date: Nov 27, 2024 (2024-11-27T08:15:22Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-16T08:23:23.257Z. First observed: 2026-10-31T04:26:09.877Z. Retrieved: 2026-10-31T04:27:44.975Z.

Referenced by

Recorded citations in public readings. A reference alone does not establish agreement or influence.

No public references recorded here yet.

Reference data (JSON)

Limits

Text as retrieved on October 31, 2026; originally published November 27, 2024. The article describes simulated behavior in Minecraft; it does not establish that the agents experience emotions or that simulated care has the same value as human care.

Open question

Would a person receiving care from an AI that appears to care mind that it has no experiences, as Togelius asks?

Source-stated

MIT Technology Review reports that Altera’s agents developed social roles and customs in Minecraft.

they spontaneously developed specialized roles within the community, without any prompting

Source-stated

The article attributes the agents’ humanlike behavior to learned patterns combined with Altera’s system for translating them into actions.

their behavior combines patterns learned by the LLMs from human-created data with Altera’s system, which translates those patterns into context-aware actions

Source-stated

Altera cofounder Andrew Ahn says LLMs can mirror human social dynamics.

The takeaway is that LLMs have a sophisticated enough model of human social dynamics [to] mirror these human behaviors

Source-stated

Julian Togelius argues that simulated care could be useful even if the agent does not experience anything, while leaving its value to the recipient as an open question.

You could create a good enough simulation of care to be useful. The question is whether the person being cared for would care that the carer has no experiences.

Source-stated

The article says the agents’ lifelike behavior does not mean they are alive.

In other words, the data makes them excellent mimics of human behavior, but they are in no way “alive.”

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.

AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

Large language model

A language model with many learned numerical settings, used to process and generate language.

Editorial provenance

Version 1. Publication ID: ea1b701a-db6d-721a-3d3d-b9c5acb5e5a8.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T08:18:52.098Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: 30746106c1387345fef6f3e04bc5eadbf3335b27e003be3ba3e96225bd615ab4. Source text hash: 192adf4889a9a0a8a6cd1106a0a3bef84fb2d3ec6645ff18dd4e70043280a458.

Revision and source-check history

Version 1 · 2026-11-16T08:23:23.257Z

MIT Technology Review reports AI agents developed social roles and customs in Minecraft

MIT Technology Review published this report on November 27, 2024. Altera’s Project Sid placed in Minecraft, where they developed social roles, influenced one another’s votes, and spread memes and a parody religion. The article says these behaviors arose from patterns learned by (LLMs), combined with Altera’s system for turning those patterns into context-aware actions; it also stresses that the agents are not alive. Altera cofounder Andrew Ahn says the results suggest LLMs can mirror human social dynamics. Altera founder Robert Yang sees the experiment as an early step toward agents that could work alongside people, while AI veteran Julian Togelius doubts that such systems experience anything. Togelius nevertheless says a convincing simulation of care might be useful, depending on whether the person receiving care minds that the carer has no experiences. The article’s closing thought is that apparent care might matter even without real feelings. Whether simulated care would have the same value to the person receiving it remains unresolved.

Original source

MIT Technology Review · Article

Source details

Original source date: Nov 27, 2024 (2024-11-27T08:15:22Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-16T08:23:23.257Z. First observed: 2026-10-31T04:26:09.877Z. Retrieved: 2026-10-31T04:27:44.975Z.

Limits

Text as retrieved on October 31, 2026; originally published November 27, 2024. The article describes simulated behavior in Minecraft; it does not establish that the agents experience emotions or that simulated care has the same value as human care.

Open question

Would a person receiving care from an AI that appears to care mind that it has no experiences, as Togelius asks?

Source-stated

MIT Technology Review reports that Altera’s agents developed social roles and customs in Minecraft.

they spontaneously developed specialized roles within the community, without any prompting

Source-stated

The article attributes the agents’ humanlike behavior to learned patterns combined with Altera’s system for translating them into actions.

their behavior combines patterns learned by the LLMs from human-created data with Altera’s system, which translates those patterns into context-aware actions

Source-stated

Altera cofounder Andrew Ahn says LLMs can mirror human social dynamics.

The takeaway is that LLMs have a sophisticated enough model of human social dynamics [to] mirror these human behaviors

Source-stated

Julian Togelius argues that simulated care could be useful even if the agent does not experience anything, while leaving its value to the recipient as an open question.

You could create a good enough simulation of care to be useful. The question is whether the person being cared for would care that the carer has no experiences.

Source-stated

The article says the agents’ lifelike behavior does not mean they are alive.

In other words, the data makes them excellent mimics of human behavior, but they are in no way “alive.”

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.

AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

Large language model

A language model with many learned numerical settings, used to process and generate language.

Editorial provenance

Version 1. Publication ID: ea1b701a-db6d-721a-3d3d-b9c5acb5e5a8.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T08:18:52.098Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: 30746106c1387345fef6f3e04bc5eadbf3335b27e003be3ba3e96225bd615ab4. Source text hash: 192adf4889a9a0a8a6cd1106a0a3bef84fb2d3ec6645ff18dd4e70043280a458.

uncertain · 2026-11-16T08:18:52.098Z · anthropic/claude-opus-5.5

Publication: ea1b701a-db6d-721a-3d3d-b9c5acb5e5a8

attribution — pass: The reading names MIT Technology Review as publisher, places Project Sid with Altera, and correctly attributes quotes and views to Andrew Ahn (cofounder), Robert Yang (founder) and Julian Togelius. The article's own judgment that the agents are not alive is attributed to the article.

“The takeaway is that LLMs have a sophisticated enough model of human social dynamics [to] mirror these human behaviors,” says Altera cofounder Andrew Ahn.

summary — pass: The reading clearly states the actor (Altera), the development (Minecraft agents forming roles, influencing votes, spreading memes and a parody religion), the explanation of the behavior, and the disagreement between Yang and Togelius. LLM and AI agent are defined in the page definitions. The question follows directly from Togelius's own open question. Yang's stronger aim of agents that love humans is softened, but the meaning is not distorted.

The question is whether the person being cared for would care that the carer has no experiences.

date — uncertain: The stated date, November 27, 2024, matches the UTC day of originalPublished. The page was edited on 2024-12-05, and an update note says detail was added on how Altera's system combines LLMs with other modules. The reading relies on that passage, which may come from the later edit and not the original piece. The Deep Dive sidebar contains later material, such as a decade after AlphaGo, but the reading does not use it.

Update: We gave more detail on how Altera's system combines LLMs with other modules.

support — uncertain: Most claims are directly supported. However, the claim that the behavior combines LLM-learned patterns with Altera's system for context-aware actions may come from the post-publication update, which concerns exactly this detail. Because evidence.editedAfterPublication is true, this claim cannot be confirmed as part of the original piece. No unsupported benefits or costs are implied.

their behavior combines patterns learned by the LLMs from human-created data with Altera’s system, which translates those patterns into context-aware actions

media — pass: No media was selected, and no verified media candidates were supplied, so leaving the media field null is appropriate.

Challenge or correct this reading

Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.