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?
79 72 85 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
80 72 88 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 1 promise resolved, 5 needed for a score.
All types combined: 79 (likely 74 to 84, high confidence), from 102 claims: 64 factual claims, 38 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Gary Marcus's claims
42 claims
“Also dear to my heart, Bengio et al echoed the fundamental theme of this blog:”
Gary Marcus, author of Marcus on AI 26 October 2023 A CERN for AI and the Global Governance of AI (Marcus on AI)
Bengio et al.'s proposal states that research breakthroughs are needed to solve some of today's technical challenges in creating AI with safe and ethical objectives.
Resolved true The record shows it happened or was so.
What would count as true or false
Judged as of 2023-10-26. True: the Bengio et al. proposal states, in substance, that research breakthroughs are needed to solve some current technical challenges in building AI with safe and ethical objectives, e.g. language saying such challenges cannot be solved simply by scaling or with existing methods. Mostly true: the paper makes this point, but in somewhat different wording or scope, e.g. it covers safety but not ethics explicitly. Mixed: the paper calls for more safety research but does not clearly say that breakthroughs are needed. Mostly false: the paper only touches on the theme, and Marcus's 'echo' overstates it. False: the paper contains no such statement or says the opposite.
Evidence
- Managing extreme AI risks amid rapid progress arXiv (authors: Bengio et al.) · 15 September 2023
“There are many open technical challenges in ensuring the safety and ethical use of generalist, autonomous AI systems. Unlike advancing AI capabilities, these challenges cannot be addressed by simply using more computing power to train bigger models.”
- Managing extreme AI risks amid rapid progress arXiv (authors: Bengio et al.) · 15 September 2023
“A first set of R&D areas needs breakthroughs to enable reliably safe AI.”
- Managing extreme AI risks amid rapid progress arXiv (authors: Bengio et al.) · 15 September 2023
“In some cases, leaps of progress may be needed; we thus do not know if technical work can fundamentally solve these challenges in time.”
Reasoning
The Bengio et al. paper says there are open technical challenges in ensuring the safety and ethical use of generalist, autonomous AI systems, and that these cannot be addressed simply by using more computing power to train bigger models (E1). It also says that a first set of R&D areas needs breakthroughs to enable reliably safe AI (E2), and that leaps of progress may be needed in some cases (E3). Together these state in substance that breakthroughs are needed for current safety and ethics challenges, which meets the 'true' criterion.
Cite this claim
Gary Marcus, “Bengio et al.'s proposal states that research breakthroughs are needed to solve some of today's technical challenges in creating AI with safe and ethical objectives.” (said 26 October 2023). Verdict: resolved true. Reliability API, claim c_37c8e2357083, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_37c8e2357083
Claims without a verdict
Is a verdict here wrong, or do you speak for Gary Marcus? 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.