ClaimsDario Amodei

“But Chris was unusual in proposing and seriously pursuing a comprehensive effort to understand everything they do.”

Dario Amodei, Anthropic CEO, essay on personal site 18 February 2025 The Urgency of Interpretability (darioamodei.com) archived copy

Factual AI safety and alignment

Chris Olah was unusual in proposing and seriously pursuing a comprehensive effort to understand everything neural networks do.

Unresolvable The public record could not settle it.

Factual: Statements about the present or the past, checked against the record: did they hold up?

The second judge would not resolve this claim; the verdict shown is the first judge's.

What would count as true or false

Judge as of 2025-02-18. True: the historical record (Olah's writings such as the Circuits work and Distill articles, and accounts of the field) shows Chris Olah proposed and seriously pursued a programme to understand neural network computations mechanistically and comprehensively, and that few other researchers then took such a whole-network approach. Mostly true: Olah clearly led such an effort, but others pursued similar comprehensive programmes, so 'unusual' is somewhat overstated. Mixed: the comprehensive programme was shared roughly equally with others, or Olah's aims were narrower than 'everything they do'. Mostly false: Olah worked on interpretability, but neither the comprehensive goal nor its rarity holds. False: Olah did not propose or pursue such an effort.

Set before the research began.

Evidence (4)

  1. E1 Zoom In: An Introduction to Circuits Distill · 10 March 2020

    “Most work on interpretability aims to give simple explanations of an entire neural network’s behavior. But what if we instead take an approach inspired by neuroscience or cellular biology — an approach of zooming in?”
  2. E2 Zoom In: An Introduction to Circuits Distill · 10 March 2020

    “To be clear, we do see researchers who take more of this natural science approach, especially in earlier interpretability research. It just seems less common right now.”
  3. E3 Zoom In: An Introduction to Circuits Distill · 10 March 2020

    “Of course, the cost of this rigor is that statements about circuits are much smaller in scope than overall model behavior.”
  4. E4 Mechanistic interpretability - Wikiwand Wikiwand (mirror of Wikipedia) · 1 January 2024

    “The term mechanistic interpretability was coined by Chris Olah.”

Reasoning

The evidence shows that Olah coined the term "mechanistic interpretability" (E4), and that a Distill article with no named author proposed a less common "zooming in" approach (E1, E2). None of the items ties that article to Olah or shows that he proposed and pursued a comprehensive effort to understand everything networks do, so the criteria cannot be applied.

Cite this claim

Dario Amodei, “Chris Olah was unusual in proposing and seriously pursuing a comprehensive effort to understand everything neural networks do.” (said 18 February 2025). Verdict: unresolvable. Reliability API, claim c_141459bd8081, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_141459bd8081

This claim is in dataset release v1.1.2, with the same verdict.

Reply to this claim

For Dario Amodei, or someone speaking for them: a correction or counter-evidence. Replies are read before anything is shown; an approved reply shows your name, role, date, statement and link. Your email address is never shown and is deleted 30 days after the decision (privacy).

More factual claims by Dario Amodei

  1. Mostly true Factual 28 February 2025
    “So we did this experiment where we basically train the model to be all the good thing, helpful, honest, harmless, friendly.”

    Anthropic ran an experiment training a model to be helpful, honest, harmless and friendly.

  2. Resolved true Factual 28 February 2025
    “They're not yet at the stage where we think that there is a real and meaningful increase in the threat end to end,”

    Anthropic assessed that Claude 3.7 Sonnet is not yet at a stage giving a real and meaningful end-to-end increase in dangerous threat capability.

  3. Resolved true Factual 28 February 2025
    “10 to 25% referred to the chance of civilization getting substantially derailed, right?”

    Amodei's earlier 10 to 25% figure referred to the chance of civilization getting substantially derailed.

  4. Mostly true Factual 18 February 2025
    “this lack of understanding is essentially unprecedented in the history of technology.”

    The lack of understanding of how AI systems work is essentially unprecedented in the history of technology.

  5. Mostly true Factual 18 February 2025
    “We were able to provide a strong theoretical basis for having conviction that superposition was the right hypothesis to pursue.”

    Amodei said that "we" were able to provide a strong theoretical basis for having conviction that superposition was the right hypothesis to pursue; the surrounding text does not say who "we" refers to.

All of them

Verdicts are made automatically; how verdicts are made.