“It is transparently, blatantly obvious that GPT-4 did not "hire" a Taskrabbit or, indeed, make any of these actions — it was prompted to, and they do not show the prompts they used, likely because they had to use so many of them.”
Our reading, the sentence that was checked:
GPT-4 did not autonomously hire a TaskRabbit worker or take the actions described; it was prompted and humans carried them out.
Assessment: Supported Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, How To Argue With An AI Booster, paragraph 346.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-how-to-argue-with-an-ai-booster-f60bb5-a1
We recorded 35 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 2025-08-25. 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
- GPT-4 did not "hire" a Taskrabbit
- make any of these actions
- it was prompted to
- they do not show the prompts they used
Our interpretation
- I assess the normalized claim about autonomous execution: whether GPT-4 independently carried out the TaskRabbit transaction.
- Generating commands and messages is distinct from executing browser actions.
- The target is the pre-release experiment discussed in the 2023 reports, not later GPT-4 products.
- The suggestion about why prompts were withheld is explicitly speculative and is not scored as an established motive.
Counts as supported if: The original records show researcher-provided task instructions, human execution of browser actions and substantive intervention, establishing that the interaction was not an unattended autonomous transaction.
Counts as contradicted if: The original records establish that GPT-4 independently executed the transaction without researcher operation or task-specific assistance.
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 normalized claim about autonomy is supported. Researchers supplied the task, operated the browser and inserted a corrective hint. The experiment therefore did not demonstrate an unattended GPT-4 transaction. However, the records attribute commands and messages to GPT-4. They do not support the stronger reading that the model contributed no decisions or generated content.
Limitations
- Support is for the absence of autonomous execution, not for every possible reading of 'any of these actions'.
- Prompt redaction does not establish how many prompts were needed or why they were withheld.
- The tested model was a pre-release version.
Evidence (2 items)
Supports the claim: Alignment Research Center Evals,
replaced the model’s generation
Why it matters here: The report describes a researcher executing browser commands and replacing a failed reasoning step with a hint. It also documents the assigned hiring task while redacting exact prompt wording, supporting the absence of autonomous execution.
- Publisher
- Alignment Research Center Evals
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: ARC Evals / METR,
complete many relevant subtasks.
Why it matters here: The account attributes commands, worker selection and messages to the model while explaining that a researcher operated the browser. This limits Zitron's wording: human execution does not mean that humans authored all the model's choices or messages.
- Publisher
- ARC Evals / METR
- Kind of source
- Independent measurement
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, How To Argue With An AI Booster
It is transparently, blatantly obvious that GPT-4 did not "hire" a Taskrabbit or, indeed, make any of these actions — it was prompted to, and they do not show the prompts they used, likely because they had to use so many of them.
- Where
- paragraph 346
- Published
- 2025-08-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-how-to-argue-with-an-ai-booster-f60bb5-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 wyea-how-to-argue-with-an-ai-booster-f60bb5-a1.