Gary Marcus

Rank 13 of 47
|
Score 83
3/8/2025, 5:38:38 PM
X
In reply to:
Sabine Hossenfelder
@skdh
·
312d
@GaryMarcus @tegmark I agree but I think there are cases where the step from the first (data analysis and search etc) to the latter (a Nobel worthy discovery) is so small that AIs can now almost close it on their own.
Gary Marcus
@GaryMarcus
·
312d
@skdh @tegmark Maybe we are talking past each other. I don’t doubt that a scientist with an eye of what they want can use these things as tools.

There is a difference between a machine conceiving of an executing a research program that is original and important and Nobel worthy, and a Nobel…
Sabine Hossenfelder
@skdh
·
312d
@GaryMarcus @tegmark the pattern isn't complex, you just need to be able to chew through a lot of data. See, if this was possible in 2019, how much will be possible in 2025:

https://www.nature.com/articles/…
Gary Marcus
@GaryMarcus
·
312d
@skdh @tegmark 1. Sure there is a lot left to discover.
2. But the causal reasoning is too poor, the evaluation of complex patterns is too unreliability, and the originality too low.
3. Best uses cases for quite some time will be things like AlphaFold where a scientist uses AI as a tool, and…
Sabine Hossenfelder
@skdh
·
312d
@GaryMarcus @tegmark Sorry @GaryMarcus I am with Max on this. There are almost certainly "Nobel quality discoveries" hidden in the published scientific literature because no human has an overview. AI can almost do this now, and these discoveries are no less "original" than what most scientists…
Gary Marcus
@GaryMarcus
·
312d
@tegmark Sorry @tegmark, with all due respect, I think that it is you that is mistaken.

Black box AI is not going (on its own) going to make original Nobel quality discoveries any time soon.

Roughly speaking, Dario is making an extrapolation that looks like this:

The statement engages in a discussion about the capabilities of AI in scientific discovery, questioning the originality and significance of AI's contributions compared to historical achievements. It reflects on AI's potential to identify important questions and make discoveries, contributing to the broader debate on AI's role in science.

  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 dialogue about AI's role in science. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by questioning and exploring AI's potential in scientific discovery. [+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 and dialogue, contributing to a nuanced discussion without personal attacks. [+1]