ClaimsDario Amodei

“One thing we didn’t discuss in the paper is the issue of transparency of neural net.”

Dario Amodei, OpenAI researcher, interviewee 21 July 2017 Dario Amodei on OpenAI and how AI will change the world for good and ill (80,000 Hours) archived copy

Factual AI safety and alignment

The Concrete Problems in AI Safety paper did not discuss the transparency of neural nets.

Unresolvable The public record could not settle it.

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

What would count as true or false

Test the text of 'Concrete Problems in AI Safety' (Amodei, Olah et al., arXiv 1606.06565, 2016, in the version available as of 2017-07-21). True: the paper has no section or substantive discussion of the transparency or interpretability of neural nets; at most it mentions the topic in passing (e.g. a citation or a single clause). Mostly true: the paper mentions transparency or interpretability only briefly, in a sentence or two or in a list of related areas, without treating it as a problem it discusses. Mixed: the paper deals with transparency in a limited but real way, e.g. a paragraph inside another section. Mostly false: the paper discusses transparency over several paragraphs as part of its analysis. False: the paper has a dedicated section or subsection on neural-net transparency or interpretability.

Set before the research began.

Evidence (4)

  1. E1 Transcript: Concrete Problems in AI Safety with Dario Amodei and Seth Baum - Future of Life Institute Future of Life Institute · 31 August 2016

    “There are five concrete problems: avoiding negative side effects, avoiding reward hacking, scalable oversight, safe exploration, and robustness to distributional shift.”

  2. E2 Transcript: Concrete Problems in AI Safety with Dario Amodei and Seth Baum - Future of Life Institute Future of Life Institute · 31 August 2016

    “But then, as soon as they started talking about their research, many of the issues that came up were related to things like control, and transparency, and bias.”

  3. E3 Transcript: Concrete Problems in AI Safety with Dario Amodei and Seth Baum - Future of Life Institute Future of Life Institute · 31 August 2016

    “But then, as soon as they started talking about their research, many of the issues that came up were related to things like control, and transparency, and bias.”

  4. E4 Concrete AI safety problems OpenAI · 21 June 2016 Own record

    “The authors discuss five areas:”

Reasoning

The criteria require testing the text of the paper, but no evidence item quotes or examines its full text. E1 lists five concrete problems, and transparency is not among them; E4 says only that the authors discuss five areas, without naming them. That shows only that transparency is not one of the main problems, not whether it gets a brief mention, a paragraph or related-work discussion. E2 and E3 mention transparency only as an issue that came up when 'they started talking about their research', without tying it to the paper, so the evidence does not settle which grade of the criteria applies.

Cite this claim

Dario Amodei, “The Concrete Problems in AI Safety paper did not discuss the transparency of neural nets.” (said 21 July 2017). Verdict: unresolvable. Reliability API, claim c_9890d8ecbf75, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_9890d8ecbf75

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.