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 still can't really yet go beyond asking a new novel question or a new novel conjecture or coming up with a new hypothesis that has not been thought of before.”
Demis Hassabis, Google DeepMind CEO 3 August 2025 AI could end disease, lead to "radical abundance" - 60 Minutes transcript (CBS News)
Current AI systems still can't really yet go beyond asking novel questions or conjectures, or come up with new hypotheses not thought of before.
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.
What would count as true or false
Judged as of August 2025. True if, as of then, there was no well-documented case of an AI system on its own posing a new scientific question, conjecture or hypothesis that experts recognised as truly original, and expert consensus held that AI could not yet do this. Mostly true if there were only isolated or disputed claims of such originality (for example AI co-scientist systems producing hypotheses that experts mostly saw as recombinations), with consensus broadly agreeing with the claim. Mixed if there were several credible, documented cases of AI producing truly original hypotheses or conjectures that experts accepted, but this was not general. Mostly false if such capabilities were well established in several domains by August 2025. False if AI producing new hypotheses and conjectures was routine and widely accepted by then.
Evidence
- Scientific hypothesis generation by large language models: laboratory validation in breast cancer treatment The Royal Society (Journal of the Royal Society Interface); record hosted by Oxford University Research Archive (ORA) · 4 June 2025
“In the first round of laboratory experiments, GPT4 succeeded in discovering three drug combinations (out of 12 tested) with synergy scores above the positive controls.”
- Scientific hypothesis generation by large language models: laboratory validation in breast cancer treatment The Royal Society (Journal of the Royal Society Interface); record hosted by Oxford University Research Archive (ORA) · 4 June 2025
“We conclude that LLMs are a valuable source of scientific hypotheses.”
- FunSearch: Making new discoveries in mathematical sciences using Large Language Models Google DeepMind · 14 December 2023 Own record
“This work represents the first time a new discovery has been made for challenging open problems in science or mathematics using LLMs.”
- Google DeepMind CEO predicts AGI in 5-10 years but says AI still lacks imagination The Indian Express · 22 April 2025
“They still can’t really yet go beyond asking a new novel question or a new novel conjecture or coming up with a new hypothesis that has not been thought of before,” Hassabis said in an interview that aired as part of CBS News’ 60 Minutes programme on Sunday, April 22.”
Reasoning
A peer-reviewed June 2025 study reports that GPT-4 proposed breast-cancer drug combinations, three of which beat the positive controls in lab tests. Its authors conclude that LLMs are "a valuable source of scientific hypotheses" (E1, E2). Those hypotheses are combinations of existing drugs. The evidence doesn't show experts accepting them as truly original conjectures. DeepMind's FunSearch claim of a new mathematical discovery (E3) comes from Hassabis's own organisation, so it counts only as an isolated, self-reported example. Before August 2025 there were some isolated, documented cases of AI generating hypotheses, but nothing showing the ability was established or widely accepted, which fits "mostly true" under the criteria.
Cite this claim
Demis Hassabis, “Current AI systems still can't really yet go beyond asking novel questions or conjectures, or come up with new hypotheses not thought of before.” (said 3 August 2025). Verdict: mostly true. Reliability API, claim c_9a78d74a9a0c, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_9a78d74a9a0c
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.