ClaimsRodney Brooks

“Oh, this is already happening”

Rodney Brooks, blogger 1 January 2018 My Dated Predictions (rodneybrooks.com) archived copy

Factual AI capabilities

Academic rumblings about the limits of Deep Learning are already happening as of 2018.

Resolved true The record shows it happened or was so.

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 2018-01-01. True: by the end of 2017 or start of 2018, published academic work or prominent academic commentary criticised or pointed to limits of deep learning (for example Gary Marcus's 'Deep Learning: A Critical Appraisal', posted January 2018, or earlier papers and talks by researchers on its limits). Mostly true: such criticism existed but was mainly from a few individuals or appeared just around the claim date. Mixed: critiques existed only outside academia (press, industry), with little academic content. Mostly false: only isolated or tangential remarks existed, overstated as a trend. False: there is evidence that no such academic critique of deep learning's limits existed by early 2018.

Set before the research began.

Evidence (4)

  1. E1 Deep Learning: A Critical Appraisal arXiv (author: Gary Marcus, New York University)

    “Yet deep learning may well be approaching a wall, much as I anticipated earlier, at beginning of the resurgence (Marcus, 2012), and as leading figures like Hinton (Sabour, Frosst, & Hinton, 2017) and Chollet (2017) have begun to imply in recent months.”
  2. E2 Failures of Gradient-Based Deep Learning PMLR (Proceedings of Machine Learning Research), ICML 2017 · 17 July 2017

    “We describe four types of simple problems, for which the gradient-based algorithms commonly used in deep learning either fail or suffer from significant difficulties.”
  3. E3 An Interview with François Chollet DataCamp · 18 December 2017

    “One important limitation of our capabilities on that front is that our models can only handle inputs that are extremely close to what they've seen before -- you can't get very far from your training data.”
  4. E4 Predictions Scorecard, 2019 January 01 rodneybrooks.com (Rodney Brooks) · 1 January 2019 Own record

    “There were plenty of papers published on limits of Deep Learning.”

Reasoning

E2 is a peer-reviewed ICML 2017 paper, published July 2017, that describes simple problems on which the gradient-based methods common in deep learning "either fail or suffer from significant difficulties". That is academic work on the limits of deep learning from before the claim date. E1 is Gary Marcus's critical appraisal, the example named in the resolution criteria, and it cites recent remarks on deep learning's limits by Hinton (Sabour, Frosst & Hinton, 2017) and Chollet (2017). In E3, an interview from December 2017, Chollet says models cannot get far from their training data. Taken together, the independent sources show academic criticism of deep learning's limits existed by the start of 2018, so the claim resolves true.

Cite this claim

Rodney Brooks, “Academic rumblings about the limits of Deep Learning are already happening as of 2018.” (said 1 January 2018). Verdict: resolved true. Reliability API, claim c_125b965f7967, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_125b965f7967

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

Reply to this claim

For Rodney Brooks, 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 Rodney Brooks

  1. Mostly true Factual 1 January 2026
    “But since 1991 I have made a distinction between two concepts where a machine, or creature can be either, neither, or both situated and embodied.”

    Since 1991 Brooks has distinguished between situatedness and embodiment as two separate properties a machine or creature can have.

  2. Mostly true Factual 1 January 2026
    “Once the pick orientation and vacuum gripper selection has been made it is really an open loop as all the work is done by the indiscriminate force of suction and the mutual compliance of the gripper…”

    Once pick orientation and vacuum gripper are selected, suction grasping is open loop, relying on suction force and compliance of gripper and object.

  3. Mostly true Factual 1 January 2025
    “Each column from bottom to top is a pure feed forward network, with no search, no iteration, no conventional algorithm at all.”

    In the 2017 'Attention Is All You Need' transformer architecture, each column from bottom to top is a pure feed-forward network with no search, iteration or conventional algorithm.

  4. Mostly true Factual 1 January 2024
    “By March of 2023, it was clear that the next big thing had arrived in AI, and that it was Large Language Models.”

    By March 2023 it was clear that large language models were the next big thing in AI.

  5. Resolved true Factual 18 April 2023
    “@yejinchoinka is providing a “common sense” rebuttal on the @TED stage.”

    Yejin Choi gave a "common sense" rebuttal on the TED stage in April 2023.

All of them

Verdicts are made automatically; how verdicts are made.