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?
76 69 82 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
56 42 70 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 4 promises resolved, 5 needed for a score.
All types combined: 72 (likely 66 to 78, high confidence), from 93 claims: 76 factual claims, 17 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Dario Amodei's claims
49 claims
“It's not anywhere near up to the task yet.”
Dario Amodei, Anthropic CEO 8 August 2023 Dario Amodei (Anthropic CEO) - $10 Billion Models, OpenAI, Scaling, & AGI in 2 years (Dwarkesh Podcast)
As of August 2023, mechanistic interpretability is not nearly capable enough to verify that a model is aligned.
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.
The evidence is the person's or their organisation's own records.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judge as of August 2023. True: the state of the field at that time (published interpretability research, expert surveys and assessments, including from Anthropic, DeepMind and independent researchers) shows that mechanistic interpretability could not verify alignment of frontier-scale models, i.e. it was far from that capability. Mostly true: the field clearly could not do this, but some evidence shows meaningful partial capability (for example verification of limited properties in small models) that makes 'not anywhere near' a slight overstatement. Mixed: credible expert assessments of the time were split roughly evenly on whether interpretability was close to verifying alignment. Mostly false: contemporaneous evidence shows interpretability could already verify significant alignment-relevant properties of real models, so the claim greatly understates progress. False: evidence shows that interpretability methods in August 2023 could verify that frontier models were aligned.
Evidence
- arxiv-2209.11895.md HuggingFace dataset page summarising Olsson et al., 'In-context Learning and Induction Heads' (arXiv:2209.11895) · 2022 Own record
“For small attention-only models, we present strong, causal evidence; for larger models with MLPs, we present correlational evidence”
- arxiv-2209.11895.md HuggingFace dataset page summarising Olsson et al., 'In-context Learning and Induction Heads' (arXiv:2209.11895) · 2022 Own record
“For 2-layer attention-only models, we were able to show precisely that induction heads implement this pattern copying behavior.”
- Superposition, memorization, and double descent Anthropic · 5 January 2023 Own record
“Despite overfitting being a central problem in machine learning, we have little mechanistic understanding of what exactly is going on when deep learning models overfit or memorize examples.”
- Towards monosemanticity Anthropic · 5 October 2023 Own record
“we have built machinery that lets us find these units in small transformer models.”
Reasoning
Anthropic-linked research from around that time reports its strongest mechanistic results on small models, with only correlational evidence for larger ones [E1][E4]. E1 says causal evidence covered only small attention-only models, and evidence for larger models was only correlational. E3 notes that the field had "little mechanistic understanding" of basic phenomena such as memorization, and E4 (October 2023) still describes the tools as working in "small transformer models". E2 shows real partial capability, a precise mechanistic explanation of one behaviour in 2-layer models, which fits the criteria's "mostly true" case, though nothing in the evidence suggests that alignment of frontier-scale models could be verified.
Cite this claim
Dario Amodei, “As of August 2023, mechanistic interpretability is not nearly capable enough to verify that a model is aligned.” (said 8 August 2023). Verdict: mostly true. Reliability API, claim c_5787ac923a7b, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_5787ac923a7b
Claims without a verdict
- Unresolvable 157
- Not yet due 40
- Own records only 43
- Not researched 134
- Not checkable 864
Is a verdict here wrong, or do you speak for Dario Amodei? 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.