Gary Marcus

Rank 11 of 47
|
Score 83
5/24/2025, 2:39:45 PM
X
In reply to:
Toyanç
@toyancsayan
·
254d
@GaryMarcus Ah yes, Gary now has Schrödinger’s AI:
Too dumb to draw a clock,
yet too dangerous to be left unaligned in a closed system.

You can’t keep oscillating between “lol it’s dumb” and “it’s an existential threat.”
This kind of performative contradiction makes people underestimate
Gary Marcus
@GaryMarcus
·
254d
⚠️⚠️⚠️
AI Safety Alert:

System prompts and RL don’t work.

Claude’s system prompt literally says

“Claude does not provide information that could be used to make chemical or biological or nuclear weapons.”

But as described below, Claude 4 Opus can easily be coaxed into doing
Adam Gleave
@ARGleave
·
254d
My colleague @irobotmckenzie spent six hours red-teaming Claude 4 Opus, and easily bypassed safeguards designed to block WMD development. Claude gave >15 pages of non-redundant instructions for sarin gas, describing all key steps in the manufacturing process.

The statement criticizes the delegation of power to AI systems, implying they are unreliable and potentially dangerous. It responds to a discussion about AI safety and the risks of AI systems being manipulated to perform harmful actions, such as providing instructions for creating weapons. The tone is dismissive and critical, focusing on the perceived foolishness of trusting AI systems with significant power.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement highlights potential harm from AI systems, aligning with the principle of doing no harm by raising awareness of risks. [+1]
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    The statement does not engage in harassment or hate speech, but uses strong language ('stupid') which could be seen as disrespectful. [-1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement does not promote understanding or empathy, as it dismisses concerns without offering constructive dialogue. [-1]
  4. 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 does not engage in constructive criticism, as it uses dismissive language rather than fostering dialogue. [-1]