Open letter responding to Yann LeCun
We recorded 5 claims from this post. 2 were checked against evidence. 3 can’t be checked. 2 more are held. Read the post at its source.
Claims from this post (5)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: UnresolvedSaid
Observation
“In 2019 I pointed out through a series of examples on Twitter with GPT-2 that LLMs were subject to essentially the same critique: greedy with respect to data, uninterpretable, unreliable as reasoners, subject to hallucinations, etc, that they lacked world models (“they don't develop robust representations of *how events unfold over time*), and that they would be inadequate on their own.”
Our reading: In 2019, using a series of GPT-2 examples on Twitter, Marcus pointed out that LLMs were greedy with respect to data, uninterpretable, unreliable as reasoners, and subject to hallucinations, that they lacked world models, and that they would be inadequate on their own.
No horizon Evidence as of
Limitation: The reply does not document the boundaries or full contents of the claimed 2019 GPT-2 Twitter series [e1].
Assessment: UnresolvedSaid
Observation
“Nowadays we both (unlike many other people) stress the absolutely essential nature of common sense, physical reasoning and world models, and the failure of current architectures to handle those well.”
Our reading: Marcus and LeCun, unlike many other people, now both stress that common sense, physical reasoning and world models are absolutely essential, and that current architectures fail to handle them well.
No horizon Evidence as of
Limitation: Public agreement does not establish who originated an idea or whether one speaker changed his position.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“To take a different kind of example, I have long argued that models need to innately be more structured than pure multilayer perceptions, even prior to learning.”
Our reading: AI models need more innate structure than pure multilayer perceptrons, even before learning.
No horizon
Qualifier: prior to learning
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Of course my biggest (and most difficult to confirm or reject) positive conjectures, all made in 2001 in The Algebraic Mind, remain speculative: the importance of representing kinds distinctly from individuals, separating reasoning from latent representations, incorporating explicit compositional structure and explicit operations over variables, incorporating some degree of innate pre learning, and exploiting explicit knowledge databases. I stand by all of them.”
Our reading: Marcus stands by his speculative 2001 conjectures, which he says are hard to confirm or reject, that it is important to represent kinds distinctly from individuals, separate reasoning from latent representations, incorporate explicit compositional structure and explicit operations over variables, incorporate some degree of innate pre-learning, and exploit explicit knowledge databases.
No horizon
Qualifier: conjectures remain speculative and difficult to confirm or reject
Assessment: Can’t be checkedSaid
Observation
“The irony of all of this is that you and I are among the minority of people who have come to fully understand just how limited LLMs are, and what we need to do next.”
Our reading: Marcus and LeCun are among the minority of people who have come to fully understand just how limited LLMs are and what needs to be done next.
No horizon
Scope: among the minority of people
Held from this post (2)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The record does not meet the current rules. Our reading, not assessed: LLMs do not provide commonsense reasoning, physical reasoning or world models for free.
- The two reviews read the claim differently