ClaimsEliezer Yudkowsky

“I previously thought that when intelligence was built, there were going to be multiple specialized systems in there.”

Eliezer Yudkowsky, interviewee 6 April 2023 Eliezer Yudkowsky - Why AI Will Kill Us, Aligning LLMs, Nature of Intelligence, SciFi, & Rationality (Dwarkesh Podcast) archived copy

Factual Their own work or organisation

Yudkowsky previously thought that building intelligence would involve multiple specialized systems.

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

A claim about Yudkowsky's own earlier views, judged from his writings before April 2023 (e.g. 'Levels of Organization in General Intelligence', 'Creating Friendly AI', Sequences posts, the Hanson-Yudkowsky FOOM debate). TRUE: those earlier writings show that he expected built intelligence to contain multiple specialised subsystems or modules. MOSTLY TRUE: the earlier writings show this view, but with qualifications. For example, he expected specialised subsystems inside an overall unified architecture, or held the view only during one period. MIXED: his earlier writings are split, endorsing specialised subsystems in some places and rejecting the idea in others to a similar extent. MOSTLY FALSE: the earlier writings mainly argue the opposite (a unified or general architecture without specialised systems), with only passing references to specialised components. FALSE: the earlier writings clearly and consistently reject the idea that intelligence would be built from multiple specialised systems.

Set before the research began.

Evidence (4)

  1. E1 Levels of Organization in General Intelligence Springer (chapter in Goertzel & Pennachin, eds., Artificial General Intelligence) · 2007 Own record

    “discusses the functional decomposition of general intelligence into a complex supersystem of interdependent internally specialized processes”
  2. E2 index.md tomrochette.com (a third-party reader's reading notes on LOGI)

    “Intelligence is a supersystem composed of many mutually interdependent subsystems - subsystems specialized not only for particular environmental skills but for particular internal functions”
  3. E3 index.md tomrochette.com (a third-party reader's reading notes on LOGI)

    “Not only is it impossible to duplicate general intelligence through the sum of (specialized) subsystems, Yudkowsky suspect it is impossible to achieve humanlike performance in most single contexts using specialized AI”
  4. E4 Debating Yudkowsky Overcoming Bias (Robin Hanson's blog) · 3 July 2011

    “In the debate, Eliezer gave six reasons to think very powerful brain architectural insights remain undiscovered:”

Reasoning

The 2007 chapter 'Levels of Organization in General Intelligence' describes general intelligence as a "complex supersystem of interdependent internally specialized processes" [E1]. Independent reading notes on the paper say the same thing: intelligence is made of many interdependent subsystems specialised for particular internal functions [E2]. The same notes also record that LOGI rejects building general intelligence as a simple sum of specialised subsystems [E3]. So in his earlier writing he expected specialised subsystems, but inside an integrated, unified architecture, which matches the qualified 'mostly true' case in the criteria.

Cite this claim

Eliezer Yudkowsky, “Yudkowsky previously thought that building intelligence would involve multiple specialized systems.” (said 6 April 2023). Verdict: mostly true. Reliability API, claim c_7e733e3ec576, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_7e733e3ec576

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. Mostly true Factual September 2025
    “I mean, as far as I know, I'm the person who originally made that argument.”

    Yudkowsky said that, as far as he knows, he was the person who originally made that argument.

  2. Resolved true Factual 7 April 2023
    “I refrained from signing because I think the letter is understating the seriousness of the situation and asking for too little to solve it.”

    Yudkowsky said he did not sign the open letter because he thinks it understates the seriousness of the situation and asks for too little to solve it.

  3. Mixed Factual 6 April 2023
    “MIRI has something called the Visible Thoughts Project, which did not get enough funding and enough personnel and was going too slowly.”

    MIRI ran a Visible Thoughts Project that did not get enough funding or personnel and was progressing too slowly.

  4. Mostly true Factual 6 April 2023
    “I have refused to deploy timelines with fancy probabilities on them consistently for many years”

    Eliezer Yudkowsky has consistently refused for many years to publish AI timelines with detailed probabilities.

  5. Resolved true Factual 31 March 2023
    “My TIME piece did not suggest nuclear strikes against countries that refuse to sign on to a global agreement against large AI training runs.”

    Yudkowsky's TIME piece did not suggest nuclear strikes against countries refusing to sign a global agreement against large AI training runs.

All of them

Verdicts are made automatically; how verdicts are made.