ClaimsRay Kurzweil

“That book proved a theorem that perceptrons couldn’t distinguish a connected figure — in which all parts are connected to each other — from a disconnected figure, something a human can do easily.”

Ray Kurzweil, book reviewer in NYT March 2017 How we'll end up merging with our technology (Kurzweil Library / NYT) archived copy

Factual AI capabilities

Minsky and Papert's book Perceptrons proved a theorem that perceptrons cannot distinguish connected from disconnected figures.

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?

What would count as true or false

True: Minsky and Papert's 'Perceptrons' (1969) contains a proven theorem showing that perceptrons, of the kind the book defines (e.g., with bounded order or diameter), cannot compute the connectedness predicate, that is, cannot distinguish connected from disconnected figures. Mostly true: the book proves this for the defined class under conditions the claim leaves out, but the core result matches. Mixed: the book discusses connectedness as a limitation, but the formal result differs materially from what is stated. Mostly false: the connectedness result is only conjectured or informally argued, or it concerns a substantially different property. False: the book contains no such result.

Set before the research began.

Evidence (4)

  1. E1 [PDF] Perceptions - Famous Deep Learning Papers Minsky and Papert, Perceptrons: An Introduction to Computational Geometry (expanded edition text, hosted at papers.baulab.info) · 1969

    “whether or not all the parts of any geometric figure are connected to one another!”

  2. E2 Perceptrons (book) - Wikipedia Wikipedia · 3 June 2008

    “showed that the order required for a perceptron to compute connectivity grew with the input size (Theorem 5.5).[18][17]”

  3. E3 Perceptrons (book) - Wikipedia Wikipedia · 3 June 2008

    “The most important one is related to the computation of some predicates, such as the XOR function, and also the important connectedness predicate.”

  4. E4 [PDF] 3 Weighted Networks – The Perceptron R. Rojas, Neural Networks, Springer-Verlag, Berlin · 1996

    “Proposition 6. No diameter limited perceptron can decide whether a geometric figure is connected or not.”

Reasoning

E2 reports that Perceptrons proved (Theorem 5.5) that the order a perceptron needs to compute connectivity grows with input size, and E3 lists connectedness among the book's main limitation results. E4 states that no diameter-limited perceptron can decide whether a figure is connected. The core result matches the claim, but it holds only for perceptrons of bounded order or limited diameter, a condition the claim leaves out.

Cite this claim

Ray Kurzweil, “Minsky and Papert's book Perceptrons proved a theorem that perceptrons cannot distinguish connected from disconnected figures.” (said March 2017). Verdict: mostly true. Reliability API, claim c_a17a2bae51c3, method 0.1.0+28da144b, dataset release v1.1.2. https://epistacklabs.com/claims/c_a17a2bae51c3

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

Reply to this claim

For Ray Kurzweil, 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 Ray Kurzweil

  1. Mostly false Factual 4 June 2026
    “That was not true a year ago. But that’s true today.”

    Large language models being roughly 50% better than human doctors at diagnosing conditions and recommending treatments was not true a year ago but is true today.

  2. Mixed Factual March 2025
    “AI hallucinations are still there, but are becoming much less frequent.”

    AI hallucinations persist but are becoming much less frequent.

  3. Mostly true Factual 29 March 2024
    “And it’s going very, very quickly — at an exponential rate”

    AI progress is advancing very quickly, at an exponential rate.

  4. Mostly true Factual 29 March 2024
    “It was not quite as good, but you could recognize it as Mozart or as Chopin.”

    Kurzweil's early computer music program produced melodies recognizable as in the style of Mozart or Chopin, though not quite as good.

  5. Mostly true Factual 26 May 2022
    “Yes, it was very much for AI.”

    The prizes received by Geoffrey Hinton and by Demis Hassabis and John Jumper were very much for AI.

All of them

Verdicts are made automatically; how verdicts are made.