“At the same talk at Stanford, Altman grandoisely asserted, with no qualification, that “we’re making AGI”, without touching on the possibility that current approaches might not get us there — despite (a) the enormous known problems with reliability, reasoning, planning and hallucinations and (b) the fact that the entire field, despite being massively well-capitalized, has struggled for over year (and in fact for decades) to solve any of these problems.”
Our reading, the sentence that was checked:
Current AI approaches have enormous known problems with reliability, reasoning, planning and hallucinations, and the entire field, despite being massively well-capitalized, has struggled for over a year, and in fact for decades, to solve any of these problems.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The Sam Altman Playbook, paragraph 4.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-sam-altman-playbook-f03177-a1
We recorded 18 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2024-05-05. 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)
Conditions
- despite being massively well-capitalized (qualifier, stated by the speaker)
- for over a year (and in fact for decades) (scope, stated by the speaker)
From the speaker
- enormous known problems with reliability, reasoning, planning and hallucinations
- the entire field
- massively well-capitalized
- for over year (and in fact for decades)
- to solve any of these problems
Our interpretation
- The ordinary reading is that substantial general problems remained unsolved, not that researchers had made no progress or solved no restricted tasks.
- The reference to decades concerns the history of general AI reliability and reasoning problems; it does not imply that modern language models existed throughout that history.
- The observation combines a current assessment with an explicitly historical duration. The target therefore ends on the post date and has no specified starting date.
- Industry capitalization is background to this technical assessment, not a claim about a particular company's finances.
Counts as supported if: Research and evaluations document substantial unresolved weaknesses across the four named areas, alongside evidence of sustained research and substantial industry investment.
Counts as contradicted if: Evidence establishes dependable general solutions to the named problems by the post date, or shows that the claimed history of research and substantial investment is materially mistaken.
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
The ordinary reading is supported. Contemporary sources document substantial reliability, reasoning, planning and hallucination limitations despite large investment. Historical research establishes that general reasoning and brittleness had resisted efforts for decades. This does not support a stronger reading that the field made no progress or lacked successful solutions to restricted problems.
Limitations
- The conclusion depends on reading solve as general resolution, rather than improvement or success on bounded tasks.
- The decades-long component is best established for general reasoning and brittleness.
- The planning paper advances a particular research position; its results should not be generalized to every architecture.
Evidence (5 items)
Supports the claim: OpenAI,
not fully reliable
Why it matters here: OpenAI explicitly acknowledged three named limitations more than a year before Marcus's post. The same section also reports reduced hallucination and improved factuality, so it supports unresolved problems rather than an absence of progress.
- Where
- Limitations
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: arXiv,
the autonomous mode
Why it matters here: The paper reports that approximately 12% of GPT-4-generated plans in the cited evaluations were executable and achieved their goals. It also discusses symbolic planners and hybrid approaches, limiting any inference that all forms of planning were unsolved.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: arXiv,
addressing and mitigating hallucinations
Why it matters here: This survey describes hallucination as a continuing deployment obstacle and reviews more than 32 mitigation techniques. It documents both persistent difficulty and active attempts to improve reliability.
- Where
- Abstract
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: Stanford HAI,
private investments in 2023
Why it matters here: The section reports $25.2 billion invested in generative AI during 2023, nearly nine times the previous year's amount. This supports substantial contemporary capitalization, without claiming that investment remained equally large throughout earlier decades.
- Publisher
- Stanford HAI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Supports the claim: Stanford University, John McCarthy,
makes them ``brittle''
Why it matters here: McCarthy's 1984 lecture describes expert systems' failures outside their intended scope and the difficulty of formalizing common-sense reasoning. The dated web version establishes a decades-long history for these general problems, although it does not document modern hallucination research over that entire period.
- Publisher
- Stanford University, John McCarthy
- Kind of source
- Secondary summary
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The Sam Altman Playbook
At the same talk at Stanford, Altman grandoisely asserted, with no qualification, that “we’re making AGI”, without touching on the possibility that current approaches might not get us there — despite (a) the enormous known problems with reliability, reasoning, planning and hallucinations and (b) the fact that the entire field, despite being massively well-capitalized, has struggled for over year (and in fact for decades) to solve any of these problems.
- Where
- paragraph 4
- Published
- 2024-05-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-sam-altman-playbook-f03177-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.