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?
83 77 88 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
69 49 87 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
69 49 87
All types combined: 81 (likely 76 to 86, high confidence), from 87 claims: 77 factual claims, 5 predictions, 5 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Sam Altman's claims
40 claims
“we decided to talk about GPT-1, 2, 3, and 4.”
Sam Altman, CEO of OpenAI, interview answer 18 March 2024 Transcript for Sam Altman: OpenAI, GPT-5, Sora, Board Saga, Elon Musk, Ilya, Power & AGI (Lex Fridman)
OpenAI chose to publicly discuss GPT-1, 2, 3 and 4 rather than build in secret until GPT-5.
Mostly true True in substance, wrong in a detail.
The second judge would not resolve this claim; the verdict shown is the first judge's.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
True: the record shows OpenAI publicly announced or discussed each of GPT-1, GPT-2, GPT-3 and GPT-4 (via papers, blog posts or releases) around the time it was developed, rather than keeping them secret until a later model. Mostly true: all four were discussed publicly, but one was disclosed only partially or after a notable delay (e.g. GPT-2 staged release, GPT-4's limited technical details). Mixed: about half of the four models were publicly discussed in a timely way. Mostly false: most were kept undisclosed until much later. False: OpenAI did not publicly discuss these models before the launch of a successor.
Evidence
- Better language models and their implications - OpenAI OpenAI · 14 February 2019 Own record
“Due to our concerns about malicious applications of the technology, we are not releasing the trained model.”
- NeurIPS Poster Language Models are Few-Shot Learners NeurIPS · 1 October 2020
“Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting.”
- GPT‑4 - OpenAI OpenAI · 14 March 2023 Own record
“We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning.”
- Awesome-LLM-papers/00_Organizations/0_OpenAI/2018 (OpenAI) (Arxiv) [GPT-1] Improving Language Understanding by Generative Pre-Training.pdf at main · guyulongcs/Awesome-LLM-papers GitHub (user repository guyulongcs/Awesome-LLM-papers)
“2018 (OpenAI) (Arxiv) [GPT-1] Improving Language Understanding by Generative Pre-Training.pdf”
Reasoning
OpenAI publicly announced GPT-2 in February 2019, but said it was "not releasing the trained model," so that disclosure was partial (E1). GPT-3 was described in a paper accepted at NeurIPS 2020 (E2), and OpenAI announced GPT-4 in March 2023 (E3). For GPT-1, the only evidence is a third-party reading list that files it as a 2018 OpenAI arXiv paper (E4); that list is weak support. Together this indicates that GPT-2, GPT-3 and GPT-4 were discussed publicly, with GPT-2 disclosed only partly, and gives weaker support for GPT-1, which the criteria grade as mostly true.
Cite this claim
Sam Altman, “OpenAI chose to publicly discuss GPT-1, 2, 3 and 4 rather than build in secret until GPT-5.” (said 18 March 2024). Verdict: mostly true. Reliability API, claim c_25512c30dd4c, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_25512c30dd4c
Claims without a verdict
Is a verdict here wrong, or do you speak for Sam Altman? 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.