“I am saying that many basic problems haven’t been solved.”
Our reading, the sentence that was checked:
Many basic problems on the path to AGI remain unsolved.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025, paragraph 42.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-o3-agi-the-art-of-the-demo-and-what-256352-a1
We recorded 17 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 2024-12-21. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- many basic problems
- haven’t been solved
Our interpretation
- In context, the ordinary reading is that several basic capability problems remain open, including reliable elementary reasoning and robust generalization to changed or unfamiliar tasks.
- A problem remaining open means that contemporary evaluations still expose systematic reliability failures. It does not mean that AI cannot solve any instances or has made no progress.
- Use the recent evaluation window identified by the surrounding text: the research discussed 'a couple months ago' through the o3 presentation 'yesterday', operationalized as October 1 through December 20, 2024.
- 'Many' is read qualitatively as multiple basic shortcomings in the discussed capabilities, not as a specified count or a majority of all AGI research problems.
Counts as supported if: Contemporaneous primary evaluations document multiple basic reliability shortcomings across elementary reasoning and adaptation to changed or unfamiliar tasks.
Counts as contradicted if: Contemporaneous evaluations establish dependable solutions across the capability problems identified in the surrounding discussion, with the cited shortcomings resolved.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
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 horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Under the contextual reading, basic reliability problems remained. GSM-Symbolic documents failures under numerical changes and irrelevant added information. The contemporary o3 evaluator reports failures on easy tasks despite a strong overall score. Together these support continuing problems in elementary reasoning and robust task adaptation. The result depends on reading 'many basic problems' through the surrounding examples, rather than as a numerical claim about all requirements for AGI.
Limitations
- The scope and meaning of 'many' are inferred and materially affect the assessment.
- The Apple experiments do not test the December o3 preview.
- The ARC page contains later updates; only its original December capability findings are used.
- Observed failures do not prove that neural networks cannot reason or that the problems are permanently unsolvable.
Evidence (3 items)
Supports the claim: arXiv,
all models declines
Why it matters here: The study documents sensitivity to changed numerical values and additional irrelevant information in elementary mathematical questions. These are observed reliability problems; the authors' explanation about the absence of genuine logical reasoning remains a hypothesis.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: ARC Prize,
fails
Why it matters here: The evaluator reports residual failures on easy tasks in the o3 preview evaluation. This supports an unresolved reliability problem in novel-task solving; the broader interpretation about human intelligence is not needed for this assessment.
- Publisher
- ARC Prize
- Kind of source
- Benchmark result
- Describes events of
- Public from
Context only: ARC Prize,
87.5%.
Why it matters here: The strong semi-private evaluation result is material evidence of progress in novel-task adaptation. It limits any interpretation that basic capabilities were absent, but does not establish that the remaining reliability problems were solved.
- Publisher
- ARC Prize
- Kind of source
- Benchmark result
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025
I am saying that many basic problems haven’t been solved.
- Where
- paragraph 42
- Published
- 2024-12-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-o3-agi-the-art-of-the-demo-and-what-256352-a1: Current. Outcome: Supported. 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-o3-agi-the-art-of-the-demo-and-what-256352-a1.