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
213 claims · page 3 of 3
“There are all these negative ways in which bad AI is either leading to fatalities or undermining democracy, et cetera,”
Gary Marcus, interviewee 20 September 2019 Gary Marcus on Why AI Needs a Reboot (Psychology Today)
Bad AI is leading to fatalities and undermining democracy.
Mostly true True in substance, wrong in a detail.
The second judge said mixed; the verdict shown is the first judge's.
What would count as true or false
Judged as of 20 September 2019. True if documented cases by then show faulty or misused AI systems contributing to deaths (e.g. the 2018 Uber self-driving fatality, Tesla Autopilot crashes) and AI-driven systems undermining democratic processes (e.g. algorithmic amplification of disinformation, micro-targeted political manipulation), as reported by credible investigations. Mostly true if both kinds of harm are documented but AI's causal role is partial or disputed in one of them. Mixed if one kind of harm is well documented and the other rests only on weak or speculative evidence. Mostly false if both rest mainly on speculation or loosely attributed links. False if there was no documented case of either by that date.
Evidence
- Senate Intel Committee Releases Bipartisan Report on Russia’s Use of Social Media | Senate Select Committee on Intelligence U.S. Senate Select Committee on Intelligence · 8 October 2019
“The new report examines Russia’s efforts to use social media to sow societal discord and influence the outcome of the 2016 election, led by the Kremlin-backed Internet Research Agency (IRA).”
- Senate Intel Committee Releases Bipartisan Report on Russia’s Use of Social Media | Senate Select Committee on Intelligence U.S. Senate Select Committee on Intelligence · 8 October 2019
“In December 2018, the Committee released two independent analyses of IRA activity, produced by New Knowledge and Graphika and the University of Oxford.”
- Senate Intel Committee Releases Bipartisan Report on Russia’s Use of Social Media | Senate Select Committee on Intelligence U.S. Senate Select Committee on Intelligence · 8 October 2019
“The Committee found IRA activity increased, rather than decreased, after Election Day 2016. Analysis of IRA-associated accounts shows a significant spike in activity after the election, increasing across Instagram (238 percent), Facebook (59 percent), Twitter (52 percent), and YouTube (84 percent).”
- Senate Intel Committee Releases Bipartisan Report on Russia’s Use of Social Media Senate Select Committee on Intelligence · 8 October 2019
“The Committee found that IRA social media activity was overtly and almost invariably supportive of then-candidate Trump to the detriment of Secretary Clinton’s campaign.”
- Tesla Model X sped up in Autopilot mode seconds before fatal crash, according to NTSB | TechCrunch TechCrunch · 7 June 2018
“The report says the Model X, while in Autopilot mode, sped up to 71 mph in the seconds leading up to the crash.”
- Safety Board Says CHP, Caltrans Lapses Played Part in Delayed Repairs Before Fatal Tesla Crash | KQED KQED · 9 September 2019
“The safety report focused on a damaged crash attenuator at the site where Apple software engineer Walter Huang, 38, smashed into a highway divider in his recently purchased Tesla Model X.”
- On Microtargeting Socially Divisive Ads: A Case Study of Russia-Linked Ad Campaigns on Facebook arXiv (ar5iv rendering); FAT* '19 conference paper
“On May 10th, 2018 the Democrats Permanent Select Committee on Intelligence released a dataset containing 3,517 Facebook advertisements”
- Facebook Political Ad Delivery Algorithms Inferred Users' Political Alignment… (2019) – AI Incident 522 MIT AI Risk Repository (AI Incident Database entry) · 1 January 2026
“Alternative interpretations: The pricing differentiation could be interpreted purely as market-driven optimization for user relevance rather than an intentional effort to polarize or restrict reach.”
Reasoning
For fatalities, a 2018 NTSB finding reported by TechCrunch says a Tesla Model X sped up while in Autopilot mode seconds before a fatal crash (E5). KQED shows the crash killed Walter Huang, though the safety board also pointed to a damaged highway barrier that had not been repaired (E6), so the AI's causal role is partial. For democracy, a FAT* 2019 conference paper studies micro-targeted, socially divisive Russia-linked Facebook ads released in 2018 (E7), and the Senate committee report says independent analyses of Internet Research Agency social-media operations meant to influence the 2016 election were already public by December 2018 (E2, E1, E4). The quotes do not show how much of that harm came from algorithms rather than people, and an AI incident entry notes other readings of Facebook's ad delivery are possible (E8). Both kinds of harm were documented before 20 September 2019, but AI's causal role is partial or disputed, which fits mostly true.
Cite this claim
Gary Marcus, “Bad AI is leading to fatalities and undermining democracy.” (said 20 September 2019). Verdict: mostly true. Reliability API, claim c_a33343db050d, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_a33343db050d
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.