Gary Marcus

Rank 13 of 47
|
Score 84
6/26/2025, 11:16:54 PM
X
In reply to:
Alex Alarga ⏹️
@AlexAlarga
·
225d
@GaryMarcus @DanHendrycks @DKokotajlo Exactly, the "superhuman narrow AI" being controllable/alignable is part of the confusion. Since we can imagine "aligning" a specialised_chess_AI, we can extrapolate that to aligning an AI "specialised at EVERYTHING". But as Dan suggested, that's AGI, not ASI.

ASI is not "just"
Gary Marcus
@GaryMarcus
·
225d
@AlexAlarga @DanHendrycks @DKokotajlo “SUPERintelligence is un-alignable by_definition” - seems like a tautological claim.

I see no reason why all smarter than human machines inherently MUST be unalignable.

but sure you can tautologically say that the subset of smarter than human intelligences that are
Alex Alarga ⏹️
@AlexAlarga
·
225d
@GaryMarcus - in that long exciting conversation with @DanHendrycks and @DKokotajlo I still didn't get to hear a clear definition of ASI, but the KEY point was made by Dan, to simplify:

"SUPERintelligence is un-alignable by definition, so we must #BanSuperintelligence on an
Machine Learning Street Talk
@MLStreetTalk
·
227d
Whether AGI arrives in 3 years or 30, all three agreed current safety measures are inadequate and political intervention may be necessary.
Watch the 2h+ deep dive on MLST. https://www.youtube.com/watch?v=…

The statement engages in a discussion about the nature of artificial superintelligence (ASI) and its potential unalignability, contributing to public discourse on AI safety and ethics.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not directly cause harm but raises concerns about the potential risks of ASI.
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by discussing the complexity and potential risks of ASI, though it could be clearer in its explanation. [+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 a critical dialogue about the nature of ASI, though it could avoid tautological reasoning to strengthen its argument. [-1]