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?
81 76 87 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 2 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
65 45 83
All types combined: 80 (likely 75 to 85, high confidence), from 84 claims: 78 factual claims, 6 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Demis Hassabis's claims
42 claims
“they certainly have modeled enough of the dynamics, put it that way, that they can pretty accurately generate whatever it is, eight seconds of consistent video”
Demis Hassabis, interviewee 23 July 2025 Transcript for Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games (#475) (Lex Fridman)
Video models such as Veo 3 can pretty accurately generate about eight seconds of video that, by eye and at a glance, looks consistent.
Mostly true True in substance, wrong in a detail.
What would count as true or false
True if, as of July 2025, Veo 3 generated clips of about eight seconds (for example 8-second outputs) that reviewers or benchmarks generally described as visually consistent and physically plausible. Mostly true if the clip length was a little different (for example 5–10 seconds) or consistency was good with minor artefacts. Mixed if eight-second clips were possible but reviewers often reported clear inconsistencies. Mostly false if Veo 3 typically could not produce coherent clips of about that length. False if Veo 3 did not generate clips of about eight seconds, or its outputs were generally incoherent.
Evidence
- AI video just took a startling leap in realism. Are we doomed? Ars Technica · 29 May 2025
“Last week, Google introduced Veo 3, its newest video generation model that can create 8-second clips with synchronized sound effects and audio dialog—a first for the company’s AI tools.”
- AI video just took a startling leap in realism. Are we doomed? Ars Technica · 29 May 2025
“Overall, Veo 3 appears to have far greater temporal coherency (having a consistent subject or theme over time) than the earlier video synthesis models we’ve tested. But of course, it’s not perfect.”
- Google Flow VEO 3: User Experience Report (May 2025) - jeffbullas.com jeffbullas.com · 2 June 2025
“Short clips vs. long narratives: VEO 3, like many generative video models, is currently optimized for short clips (around 5–10 seconds).”
- [PDF] Veo-3-Tech-Report.pdf - Googleapis.com Google DeepMind (hosted on storage.googleapis.com) · 24 August 2026 Own record
“First, while much better than previous Veo models, Veo 3 is still poor at generating text (a necessary skill for many misuse scenarios), and is generally prone to small hallucinations that mark videos as clearly fake.”
Reasoning
Ars Technica reports that Veo 3 makes 8-second clips (E1). It also says the model has far greater temporal coherency than earlier models but is "not perfect" (E2). A user report says Veo 3 is optimized for clips of about 5–10 seconds (E3). Google's own technical report says Veo 3 is still prone to small hallucinations (E4), so the clips are of the claimed length and largely consistent, with minor artefacts.
Cite this claim
Demis Hassabis, “Video models such as Veo 3 can pretty accurately generate about eight seconds of video that, by eye and at a glance, looks consistent.” (said 23 July 2025). Verdict: mostly true. Reliability API, claim c_6e4db7da3d20, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_6e4db7da3d20
Claims without a verdict
Is a verdict here wrong, or do you speak for Demis Hassabis? 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.