Escape Meridian

MIT Technology Review · Nov 16, 2023

MIT Technology Review reports Google DeepMind’s proposed AGI taxonomy

Markdown · JSON · Original source

MIT Technology Review reported on November 16, 2023, that researchers had proposed a definition and five levels of (AGI). The article says the researchers sought to reduce confusion around a term used for AI that matches or exceeds human ability across tasks. Their definition requires both breadth and high achievement, and includes learning tasks, assessing performance and asking for help when needed. The researchers say what a system can do matters more than how it works, because there is not yet enough understanding of models’ underlying processes to make those processes part of a scientifically agreed-upon definition. The article notes that the researchers regard current chatbots as emerging AGI, while saying no level beyond emerging has been achieved. It also reports that they propose ongoing evaluation if AGI is developed, and that AGI does not necessarily mean autonomy. But a sharper definition does not answer why AGI should be built: Timnit Gebru warns that the goal can be unscoped and dangerous. The article closes by welcoming clarity as a way to get past arguments over definitions and into more substantive discussion.

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 16, 2023 (2023-11-16T09:02:14Z).

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

Added to Escape Meridian: 2026-11-22T10:58:11.579Z. First observed: 2026-10-30T17:49:59.731Z. Retrieved: 2026-10-30T17:55:56.273Z.

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 30, 2026; originally published November 16, 2023. The article reports the researchers’ proposed definition and levels; it does not provide the full paper’s criteria or evidence for assessing systems against them.

Open question

How could researchers evaluate future models on an ongoing basis while reaching scientifically agreed-upon judgments about their capabilities?

Source-stated

MIT Technology Review reports that Google DeepMind researchers proposed a definition and five ascending levels of AGI, and said no level beyond emerging AGI had been achieved.

The team also outlines five ascending levels of AGI: emerging (which in their view includes cutting-edge chatbots like ChatGPT and Bard), competent, expert, virtuoso, and superhuman

Source-stated

The researchers’ definition requires both general-purpose ability and high achievement, and includes learning tasks, assessing performance and asking for assistance.

an AGI must be both general-purpose and high-achieving, not just one or the other.

Source-stated

The researchers say what an AGI can do matters more than how it does it, because the underlying processes of cutting-edge models are not sufficiently understood for the definition to focus on them.

The problem is that we don’t know enough yet about the way cutting-edge models, such as large language models, work under the hood to make this a focus of the definition.

Source-stated

The researchers suggest evaluating AGI capabilities on an ongoing basis if AGI is developed, rather than relying on a few one-off tests.

The researchers suggest that if AGI is ever developed, its capabilities should be evaluated on an ongoing basis, rather than through a handful of one-off tests.

Source-stated

The article says the researchers’ discussion does not address why AGI should be built, and reports Timnit Gebru’s warning about the endeavor.

One question the researchers don’t address in their discussion of what AGI is, is why we should build it.

Source-stated

The article closes by welcoming greater clarity as a way to move beyond arguments over definitions and toward substantive discussion.

Either way, some clarity around a long-confused concept is welcome. “Just having silly conversations is kind of uninteresting,” says Legg.

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.

Artificial general intelligence

The idea of AI with broad problem-solving, creativity and adaptability across different tasks.

Google DeepMind

Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.

Editorial provenance

Version 1. Publication ID: 0e08dd63-c646-6bd2-32dc-614c5f024f12.

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

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-22T10:54:43.132Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: 4943f942167358391400c8f2deaaa243e43adfd95a202cfb73179aaeca1b6e15. Source text hash: 3a6949ff0dcaa5eb5d48e22be26ba10ad4bb2fca53cf8b382bc6ff46e4971aa8.

Revision and source-check history

Version 1 · 2026-11-22T10:58:11.579Z

MIT Technology Review reports Google DeepMind’s proposed AGI taxonomy

MIT Technology Review reported on November 16, 2023, that researchers had proposed a definition and five levels of (AGI). The article says the researchers sought to reduce confusion around a term used for AI that matches or exceeds human ability across tasks. Their definition requires both breadth and high achievement, and includes learning tasks, assessing performance and asking for help when needed. The researchers say what a system can do matters more than how it works, because there is not yet enough understanding of models’ underlying processes to make those processes part of a scientifically agreed-upon definition. The article notes that the researchers regard current chatbots as emerging AGI, while saying no level beyond emerging has been achieved. It also reports that they propose ongoing evaluation if AGI is developed, and that AGI does not necessarily mean autonomy. But a sharper definition does not answer why AGI should be built: Timnit Gebru warns that the goal can be unscoped and dangerous. The article closes by welcoming clarity as a way to get past arguments over definitions and into more substantive discussion.

Original source

MIT Technology Review · Article

Source details

Original source date: Nov 16, 2023 (2023-11-16T09:02:14Z).

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

Added to Escape Meridian: 2026-11-22T10:58:11.579Z. First observed: 2026-10-30T17:49:59.731Z. Retrieved: 2026-10-30T17:55:56.273Z.

Limits

Text as retrieved on October 30, 2026; originally published November 16, 2023. The article reports the researchers’ proposed definition and levels; it does not provide the full paper’s criteria or evidence for assessing systems against them.

Open question

How could researchers evaluate future models on an ongoing basis while reaching scientifically agreed-upon judgments about their capabilities?

Source-stated

MIT Technology Review reports that Google DeepMind researchers proposed a definition and five ascending levels of AGI, and said no level beyond emerging AGI had been achieved.

The team also outlines five ascending levels of AGI: emerging (which in their view includes cutting-edge chatbots like ChatGPT and Bard), competent, expert, virtuoso, and superhuman

Source-stated

The researchers’ definition requires both general-purpose ability and high achievement, and includes learning tasks, assessing performance and asking for assistance.

an AGI must be both general-purpose and high-achieving, not just one or the other.

Source-stated

The researchers say what an AGI can do matters more than how it does it, because the underlying processes of cutting-edge models are not sufficiently understood for the definition to focus on them.

The problem is that we don’t know enough yet about the way cutting-edge models, such as large language models, work under the hood to make this a focus of the definition.

Source-stated

The researchers suggest evaluating AGI capabilities on an ongoing basis if AGI is developed, rather than relying on a few one-off tests.

The researchers suggest that if AGI is ever developed, its capabilities should be evaluated on an ongoing basis, rather than through a handful of one-off tests.

Source-stated

The article says the researchers’ discussion does not address why AGI should be built, and reports Timnit Gebru’s warning about the endeavor.

One question the researchers don’t address in their discussion of what AGI is, is why we should build it.

Source-stated

The article closes by welcoming greater clarity as a way to move beyond arguments over definitions and toward substantive discussion.

Either way, some clarity around a long-confused concept is welcome. “Just having silly conversations is kind of uninteresting,” says Legg.

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.

Artificial general intelligence

The idea of AI with broad problem-solving, creativity and adaptability across different tasks.

Google DeepMind

Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.

Editorial provenance

Version 1. Publication ID: 0e08dd63-c646-6bd2-32dc-614c5f024f12.

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

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-22T10:54:43.132Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: 4943f942167358391400c8f2deaaa243e43adfd95a202cfb73179aaeca1b6e15. Source text hash: 3a6949ff0dcaa5eb5d48e22be26ba10ad4bb2fca53cf8b382bc6ff46e4971aa8.

passed · 2026-11-22T10:54:43.132Z · anthropic/claude-opus-5.5

Publication: 0e08dd63-c646-6bd2-32dc-614c5f024f12

attribution — pass: The reading attributes the report to MIT Technology Review and the taxonomy to Google DeepMind researchers. It attributes the critique to Timnit Gebru and the closing remark to the article and Legg. The source is a single article, not a listing.

Now a team of Google DeepMind researchers has put out a paper that cuts through the cross talk with not just one new definition for AGI but a whole taxonomy of them.

summary — pass: The summary names the actor and the development: a definition plus five levels. It covers the key criteria, the capability-over-process rationale, current chatbots as emerging AGI, ongoing evaluation, non-autonomy and Gebru's objection. AGI is explained in the text and in the definitions. The question follows from the source's measurement discussion and is not a leading claim.

They note that no level beyond emerging AGI has been achieved.

date — pass: November 16, 2023 matches the UTC day of originalPublished, and the reading gives no finer precision. The reading uses none of the later sidebar items in the Deep Dive list, such as the AlphaGo retrospective. Forecasts and suggestions remain conditional. The page is a publisher page, not an archive copy, and dateModified is one second after publication.

support — pass: Each claim matches the source text: breadth and depth, learning, self-assessment and asking for help, and capability over process because model internals are poorly understood. It also matches ongoing evaluation, the non-autonomy point and the unaddressed question of why AGI should be built. Gebru's 'unscoped' and 'dangerous' wording is grounded in the source. No unsupported benefits or costs are implied.

Gebru noted that the hypothetical technology “sounds like an unscoped system with the apparent goal of trying to do everything for everyone under any environment.”

media — pass: No media is selected, and no verified candidates were supplied. That 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.