Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
71 64 78 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 0 predictions resolved, 5 needed for a score.
All types combined: 71 (likely 64 to 78, medium confidence), from 55 claims: 55 factual claims. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Peter Thiel's claims
4 claims
| Verdict | Kind | Said | Claim | Topic |
|---|---|---|---|---|
| Mostly false | Factual | 18 Nov 2024 | When people were writing about these things in the 18th and even 19th centuries, the idea was that the world was too big to be destroyed and that God would not do it. Factual · 18 Nov 2024 | Other global risks |
| Mostly false | Factual | 18 Nov 2024 | Thiel said he thinks the classic definition of AI was passing the Turing test, fooling people into thinking it was human 'in some sense', and that ChatGPT passed it in late 2022 or early 2023. Factual · 18 Nov 2024 | AI capabilities |
| Mostly false | Factual | 15 Jun 2012 | A Chinese man tried to launch a rocket into space in the 11th century. Factual · 15 Jun 2012 | Technology |
| Mostly false | Factual | 2007 | No real economic development took place through postwar wealth transfers, and even in the best cases the money returned to the West in Swiss accounts of Third World dictators. Factual · 2007 | Economics and markets |
“I think the classic definition of AI was something that passed the Turing test and that could fool you into thinking it was a human being in some sense that had not been passed, but was passed by ChatGPT in late 22, early 23.”
Peter Thiel, interviewee 18 November 2024 Apocalypse Now? Peter Thiel On Ancient Prophecies And Modern Tech (Hoover Institution)
Thiel said he thinks the classic definition of AI was passing the Turing test, fooling people into thinking it was human 'in some sense', and that ChatGPT passed it in late 2022 or early 2023.
Mostly false False in substance, right in a detail.
What would count as true or false
There are two parts: (a) the Turing test is the classic definition of AI; and (b) ChatGPT passed it in late 2022 or early 2023. True: peer-reviewed or widely recognised evaluations or expert consensus show that ChatGPT, or the GPT-3.5/GPT-4 models behind it in that period, passed a Turing-style test (people could not reliably tell it from a human) as of late 2022 or early 2023, and the Turing test is widely described as a classic benchmark for AI. Mostly true: a ChatGPT-family model is documented as passing a Turing-style test, but the clear evidence dates from later (for example 2024 studies of GPT-4 or GPT-4.5), or the claim's 'some sense' hedge fits only informal passing. Mixed: the evidence is split, with some informal reports of passing but formal studies showing it failed in that period. Mostly false: formal evaluations from that period show ChatGPT failing, with only anecdotes of fooling people. False: there is clear evidence it did not pass any reasonable version of the test.
Evidence
- Beyond Turing: Testing LLMs for Intelligence Communications of the ACM · 12 August 2024
“The canonical experiment to check for machine intelligence is the Turing test, proposed by Alan Turing in his 1950 paper “Computing Machinery and Intelligence.””
- Beyond Turing: Testing LLMs for Intelligence Communications of the ACM · 12 August 2024
“Large language models (LLMs) such as GPT excel at speaking like people, but have not yet convincingly passed the Turing test.”
- Does GPT-4 pass the Turing test? - ACL Anthology ACL Anthology (NAACL 2024 proceedings; authors Cameron R. Jones and Benjamin K. Bergen, UC San Diego) · June 2024
“Thebest-performingGPT-4promptpassedin49.7%ofgames,outperformingELIZA(22%)andGPT-3.5(20%),butfallingshortofthebaselinesetbyhumanparticipants(66%).”
- ChatGPT Versus Consultants: Blinded Evaluation on Answering Otorhinolaryngology Case–Based Questions PMC (National Library of Medicine), JMIR journal article · 5 December 2023
“In this study, ORL consultants easily distinguished between LLM-generated and human answers, indicating that ChatGPT failed a simplified Turing test”
Reasoning
Part (a) holds: E1 calls the Turing test the canonical experiment for machine intelligence. Part (b) does not hold. In a peer-reviewed 2024 study (E3), GPT-3.5, the model behind ChatGPT in late 2022, passed in only 20% of games, and the best GPT-4 prompt passed in 49.7%, below the 66% baseline set by human participants. A December 2023 blinded evaluation (E4) found that ChatGPT failed a simplified Turing test in a medical-question setting, and an August 2024 CACM article (E2) says LLMs had not yet convincingly passed the test, so the formal evidence does not show the claimed pass.
Cite this claim
Peter Thiel, “Thiel said he thinks the classic definition of AI was passing the Turing test, fooling people into thinking it was human 'in some sense', and that ChatGPT passed it in late 2022 or early 2023.” (said 18 November 2024). Verdict: mostly false. Reliability API, claim c_e53bf68a9c43, method 0.1.0+3e03e2df. https://epistacklabs.com/claims/c_e53bf68a9c43
Claims without a verdict
Is a verdict here wrong, or do you speak for Peter Thiel? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.