Gary Marcus

Rank 14 of 47
|
Score 83
6/5/2025, 1:25:41 PM
X
In reply to:
Benjamin Todd
@ben_j_todd
·
210d
@GaryMarcus I'm aware of those limitations – I wrote about them here: https://benjamintodd.substack.co…

But I still think it's a very useful measure.
Gary Marcus
@GaryMarcus
·
210d
@ben_j_todd please read this https://garymarcus.substack.com/…

the graph you are referring just doesn’t show what you think @ben_j_todd
Benjamin Todd
@ben_j_todd
·
210d
The length of tasks AI can do has been increasing usually fast since 2024. That's probably due to the RL reasoning model paradigm, which started then.

But Epoch estimate the current rate of scaling RL compute can't continue much beyond early 2026.

That could suggest we should

The statement is part of a public discourse on the limitations and implications of AI scaling, specifically regarding the length of tasks AI can perform. The tone is critical and seeks to engage in a deeper analysis of the situation.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and is part of a constructive critique. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by encouraging a more thorough examination of the issue. [+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 statement engages in constructive criticism without personal attacks. [+1]