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
“we built this thing called DolphinGemma, which a version of our system was trained on dolphin and whale sounds”
Demis Hassabis, CEO of Google DeepMind, interviewee 23 July 2025 Transcript for Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games (#475) (Lex Fridman)
Google DeepMind built DolphinGemma, a version of its system trained on dolphin and whale sounds.
Mostly true True in substance, wrong in a detail.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
True if, by 23 July 2025, Google DeepMind (alone or with partners such as Georgia Tech and the Wild Dolphin Project) had built DolphinGemma, a Gemma-based model trained on recordings of dolphin and whale sounds. Mostly true if DolphinGemma exists and was built by Google DeepMind but its training data were dolphin sounds only, with no whale sounds, or Google DeepMind's role was shared with others. Mixed if a model named DolphinGemma exists but Google DeepMind's involvement or the training on cetacean sounds is substantially different from the claim. Mostly false if the model was built mainly by others, with only minor involvement from Google. False if no such model existed by that date.
Evidence
- DolphinGemma: How Google AI is helping decode dolphin communication Google (blog.google) · 14 April 2025 Own record
“Trained extensively on WDP’s acoustic database of wild Atlantic spotted dolphins, DolphinGemma functions as an audio-in, audio-out model”
- DolphinGemma: How Google AI is helping decode dolphin communication Google (blog.google) · 14 April 2025 Own record
“Enter DolphinGemma. Developed by Google, this AI model makes use of specific Google audio technologies: the SoundStream tokenizer efficiently represents dolphin sounds”
- DolphinGemma: How Google AI is helping decode dolphin communication Google (blog.google) · 14 April 2025 Own record
“While trained on Atlantic spotted dolphin sounds, we anticipate its potential utility for researchers studying other cetacean species, like bottlenose or spinner dolphins.”
- DolphinGemma: How Google AI is helping decode dolphin communication | Research News Center Georgia Institute of Technology, Research News Center · 14 April 2025
“Today, on National Dolphin Day, Google, in collaboration with researchers at Georgia Tech and the field research of the Wild Dolphin Project (WDP), is announcing progress on DolphinGemma: a foundational AI model trained to learn the structure of dolphin vocalizations and generate novel dolphin-like sound sequences.”
Reasoning
Georgia Tech's independent news release (E4) confirms that by April 2025 Google had announced DolphinGemma, a foundational AI model trained on dolphin vocalizations. It describes the work as a collaboration with Georgia Tech researchers and the Wild Dolphin Project. Google's announcement (E2) says Google developed the model. It also says the model was trained on the Wild Dolphin Project's recordings of Atlantic spotted dolphins and does not mention whale sounds. So the model existed and Google had the central role in building it, but the training data appear to be dolphin-only and the work was shared with partners, which fits the 'mostly true' criterion.
Cite this claim
Demis Hassabis, “Google DeepMind built DolphinGemma, a version of its system trained on dolphin and whale sounds.” (said 23 July 2025). Verdict: mostly true. Reliability API, claim c_b666d7f80a3c, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_b666d7f80a3c
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.