ClaimsYann LeCun

“Probably not. No. But it can potentially recognize people among hundreds of millions.”

Yann LeCun, Facebook AI director, interviewee 18 February 2015 Facebook AI Director Yann LeCun on His Quest to Unleash Deep Learning (IEEE Spectrum) archived copy

Factual AI capabilities

The system would probably not be more accurate, but it can potentially recognize people among hundreds of millions.

Mostly true True in substance, wrong in a detail.

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.

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Concerns Facebook's face-recognition system (DeepFace) as of 2015-02-18. True if that system was not more accurate than humans (or than the comparison in the question) at face verification, and was able, or plausibly able, to identify people across a population of hundreds of millions (for example, Facebook's user base). Mostly true if both parts broadly hold with a small qualification (for example, accuracy roughly equal to humans). Mixed if one part holds and the other does not. Mostly false if the system was clearly more accurate and the scale claim is only loosely supported. False if both parts are contradicted.

Set before the research began.

Evidence (4)

  1. E1 Facebook's facial recognition software is now as accurate as the human brain, but what now? ExtremeTech · 19 March 2014

    “Humans, performing the same verification test on the same set of photos, scored slightly higher at 97.53%.”

  2. E2 [PDF] Closing the Gap to Human-Level Performance in Face Verification Facebook AI Research (DeepFace paper, CVPR 2014; authors Taigman, Yang, Ranzato, Wolf) · 2014 Own record

    “Our method reaches an accuracy of 97.35% on the Labeled Faces in the Wild (LFW) dataset, reducing the error of the current state of the art by more than 27%, closely approaching human-level performance.”

  3. E3 [PDF] Closing the Gap to Human-Level Performance in Face Verification Facebook AI Research (DeepFace paper, CVPR 2014; authors Taigman, Yang, Ranzato, Wolf) · 2014 Own record

    “The error rate rises to 8.7% for 4K persons with 4.4M images, showing the network scales comfortably to more persons.”

  4. E4 Facebook’s new algorithm can recognize you even if your face is hidden | Fortune Fortune · 23 June 2015

    “Using its existing algorithms, Facebook has said that it can recognize you with 98% accuracy—in fact, its software can identify you in one picture out of 800 million in less than 5 seconds.”

Reasoning

The DeepFace paper reports 97.35% accuracy on LFW, described as closely approaching human-level performance (E2), and ExtremeTech reports humans scored slightly higher at 97.53% on the same test (E1), so the system was not more accurate than humans, though nearly equal. On scale, the paper shows the network scaling to about 4,000 persons (E3), and Fortune reports Facebook's statement that its software could identify a person in one picture out of 800 million (E4), making recognition among hundreds of millions plausible. The scale support rests partly on Facebook's own statement reported after the claim date, so the verdict is mostly true.

Cite this claim

Yann LeCun, “The system would probably not be more accurate, but it can potentially recognize people among hundreds of millions.” (said 18 February 2015). Verdict: mostly true. Reliability API, claim c_5e14ac1ed01e, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_5e14ac1ed01e

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

Reply to this claim

For Yann LeCun, 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 Yann LeCun

  1. Mostly true Factual 18 March 2024
    “For starters, transformers are equivariant to permutations.”

    Transformers are equivariant to permutations, as ConvNets are equivariant to translations.

  2. Mostly true Factual 14 March 2024
    “If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.”

    There are currently no AI systems that can teach themselves to drive a car in about 20 hours of practice, as a 17-year-old can.

  3. Mostly true Factual 8 March 2024
    “The architecture is good, the architecture of the encoder is good, but the fact that you train the system to reconstruct images does not lead it to produce to long, good generic features of images.”

    Training an encoder to reconstruct images does not lead it to produce good generic image features, though the encoder architecture is good.

  4. Mostly true Factual 8 March 2024
    “LLMs can do none of those or they can only do them in a very primitive way and they don’t really understand the physical world.”

    Current LLMs can do none of understanding the physical world, persistent memory, reasoning and planning, or only in a very primitive way.

  5. Mostly true Factual 8 March 2024
    “No, there’s one thing that autoregressive LLMs or that LLMs in general, not just the autoregressive one, but including the bird style bidirectional ones, are exploiting and its self supervised…”

    LLMs, including BERT-style bidirectional ones, exploit self-supervised learning, and LeCun has been a very strong advocate of self-supervised learning for many years.

All of them

Verdicts are made automatically; how verdicts are made.