Gary Marcus

Rank 18 of 47
|
Score 63
12/20/2024, 5:58:54 PM
X
In reply to:
Cameron Williams
@Wasgo
·
212d
@GaryMarcus Here's my question of the day. If we believe that reasoning models and neurosymbolic stuff will be necessary for the future of AGI, and if we achieve those things, why would LLMs still be necessarily a part of the mix?
Gary Marcus
@GaryMarcus
·
212d
Three o3 predictions:
– People will initially be amazed
- Once they dive in, they will see that it is not reliable
- It will work best in closed-domains (like math problems) but less reliably in open-ended domains (like everyday reasoning about the real world).

The statement engages in a technical discussion about the role of large language models (LLMs) in the future of artificial general intelligence (AGI). It questions the necessity of LLMs in the context of reasoning models and neurosymbolic approaches, suggesting that while LLMs are good at aggregating distributional information, they might be replaced by more efficient methods. The tone is analytical and speculative, contributing to a broader debate on AI development.

  1. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by discussing the potential roles and limitations of LLMs in AGI, encouraging a nuanced view of AI technologies. [+1]
  2. Principle 4:
    I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
    The statement engages in constructive dialogue by responding to a question with a thoughtful analysis, avoiding personal attacks or ad hominem arguments. [+1]