“LLMs are not AGI, and (on their own) never will be; scaling alone was never going to be enough.”
Our reading, the sentence that was checked:
Scaling alone was never going to be enough to reach AGI.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, August 5, 2024, a big day in tech?, paragraph 3.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-august-5-2024-a-big-day-in-tech-b90353-2-a3
We recorded 8 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-08-05. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
Conditions
- scaling alone (qualifier, stated by the speaker)
From the speaker
- scaling alone
- was never going to be enough
- AGI
Our interpretation
- Read the sentence as an unlimited prediction that increasing the scale of LLMs alone cannot reach AGI.
- Read AGI as broadly human-level ability to learn, reason and apply knowledge across unfamiliar intellectual tasks.
- Read scaling alone as increasing model size, training data and compute while retaining the underlying method, without an essential new algorithm or external reasoning system.
- Whether inference-time computation and reinforcement learning count as scaling must be disclosed rather than silently assumed.
- 'Scaling alone' defines the qualifying development route; it is not a separate contingent antecedent.
Counts as supported if: A conclusive general result establishes that scaling the specified LLM approach cannot reach AGI. Limited gains, present failures or use of other methods do not prove necessity.
Counts as contradicted if: An independently validated AGI system is shown, through technical disclosure and suitable comparisons, to have reached that capability through scaling the existing LLM method alone.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The October 14, 2026 study links training compute to frontier performance differences among models released in 2022–2025; it does not test whether scaling alone can reach AGI [e1]. OpenAI’s September 3, 2026 disclosure describes pre-training, reinforcement learning and alignment, without isolating the contribution of scale or establishing that additional methods were necessary [e2]. The AGI Society assessment finds the published evidence insufficient to establish that Astra achieved AGI [e3]. These items establish neither AGI through scaling alone nor a conclusive general limitation on that route, leaving the forecast unresolved [e1, e2, e3].
Limitations
- Benchmark correlations do not establish whether scaling alone is sufficient for AGI [e1].
- The disclosure does not isolate scale from other development methods [e2].
- The contract leaves the AGI threshold, baseline method and treatment of reinforcement learning and inference-time compute unspecified.
- The e3 metadata gives September 14, 2026, while its displayed text gives September 10, 2026.
Evidence (3 items)
Context only: npj Artificial Intelligence,
At the frontier, 80–90% of performance differences are explained by higher training compute, implying that scale—not proprietary technology—drives frontier advances.
Why it matters here: The study provides material evidence that scale explains much observed frontier performance variation. Its analysis of models released during 2022–2025 does not establish that scaling alone reaches AGI.
- Where
- Abstract
- Publisher
- npj Artificial Intelligence
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
Context only: OpenAI,
GPT‑6 Astra brings together years of research and big bets across pre-training, reinforcement learning, and alignment.
Why it matters here: The company describes several development approaches. This disclosure does not isolate scaling alone, and use of additional approaches does not prove they were necessary.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: AGI Society,
Our current evaluation is that the published evidence does not yet support concluding that GPT-6 Astra represents the achievement Artificial General Intelligence.
Why it matters here: This assessment reviews prominent AGI declarations but finds the public evidence insufficient. It is an assessment of existing evidence, not a proof that scaling cannot reach AGI.
- Publisher
- AGI Society
- Kind of source
- Secondary summary
- Describes events of
- Public from
What would settle it
Independent validation of AGI, supported by technical disclosures and controlled comparisons establishing that scaling the existing LLM method alone produced it, would contradict the claim. A conclusive general result establishing that this route cannot reach AGI would support it.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, August 5, 2024, a big day in tech?
LLMs are not AGI, and (on their own) never will be; scaling alone was never going to be enough.
- Where
- paragraph 3
- Published
- 2024-08-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-august-5-2024-a-big-day-in-tech-b90353-2-a3: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-august-5-2024-a-big-day-in-tech-b90353-2-a3.