ClaimsYann LeCun

“The proposed AI architecture consists of multiple interconnected modules.”

Anna Dawid and Yann LeCun, authors June 2023 Introduction to Latent Variable Energy-Based Models (Les Houches lecture notes) (arXiv) archived copy

Factual AI capabilities

LeCun's proposed AI architecture consists of multiple interconnected modules.

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?

The evidence is the person's or their organisation's own records.

What would count as true or false

True if LeCun's proposed architecture as of June 2023 (the autonomous machine intelligence / JEPA-based cognitive architecture, as described in his 2022 position paper 'A Path Towards Autonomous Machine Intelligence' and his 2023 talks) is made up of several distinct modules that are connected to and exchange information with one another (for example configurator, perception, world model, cost, short-term memory, actor). Mostly true if the architecture is modular but a minor detail is off, such as some components being described as sub-parts rather than separate modules. Mixed if sources describe it as only partly modular. Mostly false if it is mainly described as a single monolithic model with only a minor modular aspect. False if the architecture is not described as made of multiple interconnected modules.

Set before the research began.

Evidence (3)

  1. E1 lecun-20230424-santa-fe-institute Scribd (uploaded copy of Yann LeCun's Santa Fe Institute slides) · 24 April 2023 Own record

    “Modular Architecture for Autonomous AI”

  2. E2 lecun-20230424-santa-fe-institute Scribd (uploaded copy of Yann LeCun's Santa Fe Institute slides) · 24 April 2023 Own record

    “Stores states and costs in short-term memory”

  3. E3 lecun-20230424-santa-fe-institute Scribd (uploaded copy of Yann LeCun's Santa Fe Institute slides) · 24 April 2023 Own record

    “Configures all other modules for the task at hand”

Reasoning

The claim concerns what LeCun's own proposal contains, so his April 2023 Santa Fe Institute slides count as the published record of that proposal. One slide is titled "Modular Architecture for Autonomous AI" [E1]. The configurator is described as configuring "all other modules for the task at hand" [E3], which shows there are several modules that act on one another. Another slide describes a short-term memory component that stores states and costs [E2], showing that information passes between components. Together these show that the architecture, as presented before June 2023, is made of several connected modules.

Cite this claim

Yann LeCun, “LeCun's proposed AI architecture consists of multiple interconnected modules.” (said June 2023). Verdict: resolved true. Reliability API, claim c_a15b3bec9d0e, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_a15b3bec9d0e

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.