“True LLMs don’t (just) memorize, and true that their models make for lousy databases (for a database a hallucination is an outright fail)”
Our reading, the sentence that was checked:
OpenAI's LLMs make lousy databases because hallucinations amount to outright failures for a database.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, This Week in AI Doublespeak, paragraph 15.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-this-week-in-ai-doublespeak-f02594-a1
We recorded 3 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 2023-03-18. 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
- their models make for lousy databases
- for a database a hallucination is an outright fail
Our interpretation
- The ordinary reading is that contemporaneous OpenAI language models could not reliably serve as standalone stores from which users retrieve accurate facts.
- This concerns factual answers generated from model parameters, not applications connected to a separate database and verification system.
- The target is the GPT-4 launch period, March 14–18, 2023: the latest public model and evaluation period before the post. This is an ongoing technical observation, not a quarterly reporting claim.
- The concession about memorization is background; the assessed proposition concerns reliability as a factual database.
Counts as supported if: Contemporaneous documentation or evaluations show that the models generated fabricated factual answers and required verification when factual reliability mattered.
Counts as contradicted if: Evidence establishes reliable factual retrieval without fabrication across ordinary factual queries for the contemporaneous models.
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
OpenAI's own launch documentation acknowledges factual fabrication and recommends verification. That supports Marcus's ordinary point that these models were unreliable substitutes for factual databases. The documented improvement over GPT-3.5 qualifies the criticism but does not contradict it.
Limitations
- This does not establish that most answers were false.
- The assessment does not cover systems that retrieve and verify information from external databases.
- The decisive source is a company disclosure rather than an independent replication.
Evidence (1 item)
Supports the claim: OpenAI,
“hallucinates” facts
Why it matters here: OpenAI acknowledged factual fabrication in the released model and recommended checks appropriate to the use case. The same section reports improved factuality over GPT-3.5, but explicitly says the problem remains.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, This Week in AI Doublespeak
True LLMs don’t (just) memorize, and true that their models make for lousy databases (for a database a hallucination is an outright fail)
- Where
- paragraph 15
- Published
- 2023-03-18
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-this-week-in-ai-doublespeak-f02594-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-this-week-in-ai-doublespeak-f02594-a1.