Nate Silver interviewed OpenAI CEO Sam Altman in about August 2022, when OpenAI was internally developing ChatGPT 3.5 and GPT-4.

The claim is hedged ('I think'), so the date need only be approximately right. True: Silver met or interviewed Sam Altman around mid-to-late 2022 (roughly July to October 2022), before ChatGPT's public release on 30 November 2022, at a time when OpenAI was developing the GPT-3.5/ChatGPT and GPT-4 models. Silver's book 'On the Edge' or published notes would establish the conversation date. Mostly true: the conversation took place in 2022 but a few months from August, or the details of what was being developed are slightly off. Mixed: the timing is substantially off (for example early 2022 or 2023) but the conversation did take place. Mostly false: the conversation took place well after ChatGPT's launch. False: no such conversation took place.

Set before the research began.

1. E1 Why Extreme Risk Takers Are Winning — With Nate Silver podscripts.co (transcript of Big Technology Podcast) · 14 August 2024 Own record

   “But Sam Altman, and I talked to him, I think, in August 2022 for the first time, that was the main interview,”

2. E2 Why Extreme Risk Takers Are Winning — With Nate Silver podscripts.co (transcript of Big Technology Podcast) · 14 August 2024 Own record

   “Right before. Yeah, so they kind of knew what was coming with 3.5 and 4, and it wasn't quite public yet.”

3. E3 GPT-4: What we (I) know about it — LessWrong LessWrong · 15 March 2023

   “According to the research statement, GPT-4 “finished training” in August of 2022.”

4. E4 GPT-4 OpenAI · 14 March 2023

   “A year ago, we trained GPT-3.5 as a first “test run” of the system.”

5. E5 Why Extreme Risk Takers Are Winning — With Nate Silver Podscripts (transcript of Big Technology Podcast) · 14 August 2024 Own record

   “But Sam Altman, and I talked to him, I think, in August 2022 for the first time, that was the main interview,”

6. E6 Why Extreme Risk Takers Are Winning — With Nate Silver Podscripts (transcript of Big Technology Podcast) · 14 August 2024 Own record

   “Yeah, so they kind of knew what was coming with 3.5 and 4, and it wasn't quite public yet.”

7. E7 GPT-4 (2022) Life Architect (Alan D. Thompson) · March 2023

   “Aug/2022 | GPT-4 finished training, available in lab.”

8. E8 GPT-4: What we (I) know about it — LessWrong LessWrong (post by Robert_AIZI, linkpost for aizi.substack.com) · 15 March 2023

   “According to the research statement, GPT-4 “finished training” in August of 2022.”

9. E9 Officially announcing On The Edge, my new book Nate Silver (natesilver.net newsletter) · 8 February 2024 Own record

   “The universe of the book “blew up” as I was roughly in the middle of the project, particularly in late 2022 surrounding the “two Sams” that feature prominently in second half of the narrative”

10. E10 Officially announcing On The Edge, my new book Nate Silver (natesilver.net newsletter) · 8 February 2024 Own record

    “Sam Altman, the CEO of OpenAI, which released GPT-3.5 — one of the most rapidly-adopted technologies of all-time — just two weeks before SBF was led off in handcuffs.”

11. E11 Officially announcing On The Edge, my new book Nate Silver (natesilver.net newsletter) · 8 February 2024 Own record

    “I spoke with accelerationists and doomers and everyone in between, including Altman — as well as people who are familiar with the technical inner workings of ChatGPT and other large language models.”

12. E12 Nate Silver charts the course of risk-taking, from Sam Bankman-Fried to Sam Altman | Fortune Fortune · 12 August 2024

    “OpenAI CEO Sam Altman said the upsides of AI are too high not to gamble on its risks, according to On the Edge.”

The only evidence for the August 2022 date is Silver's own podcast remark that he spoke to Altman 'I think, in August 2022 for the first time' (E1, E5), along with his comment that OpenAI 'knew what was coming with 3.5 and 4' (E2, E6). Silver's newsletter (E11) says he spoke with Altman, and Fortune (E12) reports Altman's views as given in On the Edge, but neither gives a date for the conversation. Independent sources (E3, E7) date the end of GPT-4 training to August 2022, which fits the development context but does not show when the conversation took place. Since the timing rests only on the subject's own statements, the claim cannot be resolved under the criteria.