Gary Marcus

Rank 15 of 47
|
Score 76
2/18/2025, 10:52:25 PM
X
In reply to:
Igor Kurganov
@IgorKurganov
·
268d
@GaryMarcus we’re still scaling…

opponents of scaling did not say “yes, we’ll have extra benefits from 100k gpus and later 1m gpus, they’ll just come from a different type of caling”
Gary Marcus
@GaryMarcus
·
268d
@IgorKurganov no, in the LLM imperialism of the last few years, it wasn’t. that was the whole point of the scaling hypothesis - “scale is all you need” meant, well, scale is all you need.

that was wrong.
Igor Kurganov
@IgorKurganov
·
268d
@GaryMarcus I should really update it to “haha, but LLMs only make up a part of it, it’s AI as a whole that’s better than humans at almost anything”

This was assumed by most people
Gary Marcus
@GaryMarcus
·
268d
@IgorKurganov also, people are already starting to enhance LLMs in various ways; LLMs will be part of the solution but unless you just redefine what an LLM is, they won’t be the whole solution.
Igor Kurganov
@IgorKurganov
·
268d
@GaryMarcus As per usual:
Igor Kurganov
@IgorKurganov
·
507d
LLM skeptics be like
Gary Marcus
@GaryMarcus
·
268d
And yet Grok 3, with 15x the compute of Grok 2, and ungodly energy bills, still stumbled on 9.11 vs 9.9 and the number of L’s in Lollapalooza.

The important thing, folks, is to always revise your timelines forward, and never back, no matter what the empirical evidence might…

The conversation involves a debate on the scaling hypothesis in AI, particularly regarding large language models (LLMs). It critiques the idea that simply increasing computational resources will lead to better AI, suggesting that other factors are necessary. The tone is somewhat sarcastic and critical, reflecting differing opinions on AI development strategies.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not directly cause harm but engages in a critical discussion about AI development.
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The conversation promotes understanding by discussing the limitations and potential of AI scaling. [+1]
  3. 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 dialogue includes sarcasm and critique but does not engage in personal attacks. [+1]