ClaimsDario Amodei

“For almost a decade now, we’ve had self-driving cars that are 99.9% safe.”

Dario Amodei, OpenAI researcher, interview answer 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 capabilities

For almost a decade before 2017, self-driving cars have existed that are 99.9% safe.

Mixed Partly true and partly false in roughly equal measure.

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

What would count as true or false

Judge as of 2017-07-21. The claim is that self-driving cars that were '99.9% safe' had existed since about 2007–2008. True: evidence shows autonomous vehicles from about 2007–2009 (e.g. DARPA Urban Challenge entrants, early Google self-driving prototypes) achieved a safety or reliability level reasonably described as 99.9%, such as completing about 99.9% of driving situations or miles without failure or intervention. Mostly true: such systems existed from slightly later (e.g. 2009–2010), or the 99.9% figure is a reasonable loose characterisation of the reliability data of the time. Mixed: capable prototypes existed for about that period, but no metric supports a 99.9% figure. Mostly false: the figure distorts a narrow result (e.g. one closed-course demonstration) into a general claim, or such systems date only from much later. False: evidence shows no self-driving cars of that era approached that reliability.

Set before the research began.

Evidence (4)

  1. E1 DARPA Urban Challenge | DARPA DARPA (U.S. Defense Advanced Research Projects Agency) · 2 January 2007

    “At 1:43 pm, Boss, the entry of the Carnegie Mellon Team, Tartan Racing, crossed the finish line first with a run time of just over four hours.”
  2. E2 The MIT – Cornell Collision and Why It Happened MIT DSpace (Springer Tracts in Advanced Robotics, The DARPA Urban Challenge) · November 2009

    “one of the first well-documented collisions between two full-size autonomous vehicles”
  3. E3 Where to? A History of Autonomous Vehicles Computer History Museum · 8 May 2014

    “The first year’s crop of entrants failed miserably, traveling barely a few miles before crashing.”
  4. E4 Where to? A History of Autonomous Vehicles Computer History Museum · 8 May 2014

    “Google’s system has guided a fleet of Prius and Lexus vehicles over half a million miles without causing any accidents, and the firm is a leading advocate for fully self-driving cars.”

Reasoning

E1 shows that capable autonomous vehicles existed by the 2007 DARPA Urban Challenge, where Carnegie Mellon's Boss finished first after a run of about four hours. E4 reports that Google's fleet had driven over half a million miles without causing accidents by 2014. However, E2 documents a collision between two autonomous vehicles in the DARPA Urban Challenge, and none of the evidence gives a metric that supports a figure of 99.9% for systems from about 2007–2008. This fits the "mixed" criterion: capable prototypes existed for about that period, but no metric supports the 99.9% figure.

Cite this claim

Dario Amodei, “For almost a decade before 2017, self-driving cars have existed that are 99.9% safe.” (said 21 July 2017). Verdict: mixed. Reliability API, claim c_91363172d61f, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_91363172d61f

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. Mixed Factual 23 September 2026
    “It’s easy to dismiss this as a one-off or curiosity, but we’ve repeatedly seen a pattern where AI performance in new intellectual domains goes from weak to superhuman in a matter of a few years.”

    AI performance in new intellectual domains has repeatedly gone from weak to superhuman within a few years.

  2. Mixed Factual 10 March 2025
    “we were measuring these trends back when models cost $1,000 or $10,000 to train”

    Amodei's group measured AI scaling trends when models cost $1,000 or $10,000 to train.

  3. Mixed Factual 28 February 2025
    “a lot of the other reasoning models in the market are trained primarily on math and competition coding,”

    Many other companies' reasoning models on the market are trained primarily on math and competition coding.

  4. Mostly true Factual 28 February 2025
    “I do think AI coding is moving the fastest of all the other areas.”

    Amodei said he thinks AI coding is moving the fastest of all the areas of AI.

  5. Mostly true Factual 28 February 2025
    “Because when it knows that it doesn't get any gain out of thinking further, it doesn't think for longer.”

    The model sometimes gives a very short response because, when it knows it gains nothing from thinking further, it does not think for longer; a bound on how long it thinks is still described as valuable.

All of them

Verdicts are made automatically; how verdicts are made.