“There’s not a huge moat here to protect OpenAI from competitors. The central technology at OpenAI (as far as anyone knows) is large language models, and many other companies know how to build them.”
Our reading, the sentence that was checked:
There is not a huge moat to protect OpenAI from competitors because its central technology, as far as anyone knows, is large language models, which many other companies know how to build.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Top 5 reasons OpenAI is probably not worth 90 billion dollars, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-top-5-reasons-openai-is-probably-25556c-a1
We recorded 11 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 2023-09-27. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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
- as far as anyone knows (qualifier, stated by the speaker)
From the speaker
- There’s not a huge moat
- protect OpenAI from competitors
- central technology
- large language models
- many other companies know how to build them
- as far as anyone knows
Our interpretation
- The ordinary reading is that widespread ability to build capable LLMs materially limited OpenAI's technological protection from competition at the post date.
- The statement does not require competitors to equal GPT-4 on every task; the surrounding text expressly allows a quality gap.
- The existence of rival models tests the technical premise, but does not by itself establish the size or durability of the competitive advantage.
- As far as anyone knows limits the assertion to publicly knowable technology; it is not a future conditional antecedent.
Counts as supported if: Evidence establishes that multiple independent companies could build practically competitive LLMs and that this capability materially limited any substantial technological barrier protecting OpenAI.
Counts as contradicted if: Evidence establishes that OpenAI's technology created a substantial, persistent barrier that prevented otherwise capable rivals from offering practical substitutes, or that the asserted widespread ability to build such models was absent.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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 horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The technical premise is supported: OpenAI used language modelling, and Anthropic, Google and Meta had developed rival models before the post. The broader causal conclusion remains unsettled. Rival development does not itself measure the size of a moat, while Meta's report documents a substantial capability gap. That gap is a qualification, not a direct contradiction, because Marcus allowed that competitors could be less capable.
Limitations
- Model releases demonstrate technical capability, not the absence of durable competitive advantages.
- The cited capability comparisons are developer-reported.
- No later valuation or market share is used to retroactively settle the September 2023 claim.
Evidence (4 items)
Supports the claim: OpenAI, via arXiv,
a Transformer-based model
Why it matters here: Supports the premise that a central OpenAI model used the language-modelling approach discussed in the claim. It does not measure the resulting competitive advantage.
- Where
- Abstract
- Publisher
- OpenAI, via arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: Anthropic,
We are pleased to announce Claude 2, our new model.
Why it matters here: Anthropic had released a rival model before the post. The announcement describes API and public chat access, supporting the presence of an actual competing product rather than merely a development plan.
- Publisher
- Anthropic
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: Google,
PaLM 2, our next generation language model.
Why it matters here: Google's release supports the premise that another company could develop and deploy language models. The accompanying product descriptions establish practical deployment but do not establish the absence of a large OpenAI advantage.
- Where
- Introducing PaLM 2
- Publisher
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Meta research team, via arXiv,
large gap in performance
Why it matters here: Meta's comparison both establishes another independently built LLM and documents an important capability gap. That qualifies an inference from shared technology to easy substitution, but does not contradict Marcus's express allowance that competitors might be less capable.
- Publisher
- Meta research team, via arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
Contemporaneous evidence showing whether rival models were practical substitutes and whether OpenAI-specific technical advantages materially obstructed competition, assessed against an explicit interpretation of a huge moat.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Top 5 reasons OpenAI is probably not worth 90 billion dollars
There’s not a huge moat here to protect OpenAI from competitors. The central technology at OpenAI (as far as anyone knows) is large language models, and many other companies know how to build them.
- Where
- paragraph 8
- Published
- 2023-09-27
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-top-5-reasons-openai-is-probably-25556c-a1: 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-top-5-reasons-openai-is-probably-25556c-a1.