Gary Marcus

Rank 23 of 47
|
Score 41
7/30/2025, 12:15:00 AM
X
In reply to:
Brandon
@PixelCanuck
·
21d
@GaryMarcus People will no doubt say "1.5% is really low that's great"

But in computing we talk about services that are thousands of times more reliable, like 0.00001% unreliable not 1.5% reliable
Gary Marcus
@GaryMarcus
·
21d
brutal thread: even the safest AI agent got beat 1.5% of the time.

none of this stuff is remotely secure.
Andy Zou
@andyzou_jiaming
·
21d
We deployed 44 AI agents and offered the internet $170K to attack them.

1.8M attempts, 62K breaches, including data leakage and financial loss.

🚨 Concerningly, the same exploits transfer to live production agents… (example: exfiltrating emails through calendar event) 🧵

The statement critiques the perceived inadequacy of AI security, emphasizing that a 98.5% security rate is insufficient in computing contexts where much higher reliability is expected. This reflects concerns about AI vulnerabilities and their implications.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement highlights potential harm due to AI security vulnerabilities, aligning with the principle of striving to do no harm. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    By discussing AI security issues, the statement promotes understanding and awareness of potential risks. [+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 of AI security without personal attacks. [+1]