ClaimsEliezer Yudkowsky

“Since Copycat is a model of perceptual analogy-making, rather than a model of category formation, Copycat has a limited store of preprogrammed concepts and does not learn further concepts through experience.”

Eliezer Yudkowsky, MIRI researcher, paper author 2007 Levels of Organization in General Intelligence (MIRI) archived copy

Factual AI capabilities

The Copycat program has a limited store of preprogrammed concepts and does not learn further concepts through experience.

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?

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

What would count as true or false

Judged as of 2007 against descriptions of Hofstadter and Mitchell's Copycat program. True: published descriptions show that Copycat's concepts (in the Slipnet) are a fixed, preprogrammed set and that the program does not create new concepts from experience. Mostly true: the core point holds, with a minor nuance, for example node activations and conceptual depths change during a run but no new concepts are learned. Mixed: Copycat has some limited concept learning, so the claim is about half right. Mostly false: Copycat has meaningful mechanisms for acquiring new concepts that the claim ignores. False: Copycat is documented to learn new concepts through experience as a core feature.

Set before the research began.

Evidence (4)

  1. E1 Hofstadter 1995: The copycat project: A model of mental fluidity and analogy-making Jim Davies (personal site summary of Hofstadter and Mitchell, Fluid Concepts and Creative Analogies, 1995)

    “The net changes during the course of a run, but doesn't change permanently as a result. New concepts are not created.”
  2. E2 Copycat (software) - Wikipedia Wikipedia

    “The codelets rely on an associative network, slipnet, built on pre-programmed concepts and their associations (a long-term memory).”
  3. E3 Exploring Copycat Architecture: A Cognitive Science Perspective on Analogy-Making - SlideServe SlideServe (presentation by Uma Ramamurthy, Cognitive Science Seminar) · 29 January 2003

    “Copycat’s Shortcomings • No Learning • No new Slipnet nodes can be created • No new Codelets can be generated”
  4. E4 How Could a Copycat Ever be Creative? AAAI Technical Report SS-93-01 (Douglas Hofstadter) · 1993

    “manufactures its own mutable representations”

Reasoning

E1 summarises Hofstadter and Mitchell's 1995 description of Copycat. It says the network changes during a run but not permanently, and that new concepts are not created. E2 describes the slipnet as built on pre-programmed concepts. E3 lists as a shortcoming that Copycat has no learning and that no new Slipnet nodes can be created. E4's phrase about mutable representations does not mention learning new concepts, so it does not clearly contradict the claim.

Cite this claim

Eliezer Yudkowsky, “The Copycat program has a limited store of preprogrammed concepts and does not learn further concepts through experience.” (said 2007). Verdict: resolved true. Reliability API, claim c_914c8600742f, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_914c8600742f

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

Reply to this claim

For Eliezer Yudkowsky, 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 Eliezer Yudkowsky

  1. Resolved true Factual September 2025
    “We learned that that's false, but we did not learn that you can have a thing that can redesign a tree and build an AI much smarter than itself, that builds an AI much smarter than itself, and that…”

    No AI has yet been shown to redesign a tree and build a much smarter AI that builds a still smarter AI; that scenario has not occurred.

  2. Mostly true Factual 25 March 2024
    “I was a passionate noticer of the human brain not being homogenous, and was therefore unpleasantly surprised by transformers scaling as well as they did.”

    Yudkowsky was unpleasantly surprised by how well transformers scaled, having believed the human brain is not homogeneous.

  3. Mostly true Factual 6 April 2023
    “First of all, you realize in principle that the task of minimizing losses on predicting human text does not stop when you’re as smart as a human, right?”

    In principle, the task of minimizing loss on predicting human text does not stop being improvable once a system is as smart as a human.

  4. Mostly true Factual 30 March 2023
    “Transformers are the main thing like that.”

    Transformers were the main architectural change that gave a non-linear jump in performance, compared with earlier approaches such as RNNs.

  5. Mostly true Factual 30 March 2023
    “Because the basic paradigm of machine learning that we are presently operating under is that you can have the loss function, but only for things you can evaluate.”

    The basic paradigm of machine learning currently in use is that a loss function can be defined only for things that can be evaluated.

All of them

Verdicts are made automatically; how verdicts are made.