Gary Marcus

Rank 14 of 47
|
Score 96
4/9/2024, 4:42:22 PM
X
In reply to:
The_AI_Skeptic
@The_AI_Skeptic
·
537d
@GaryMarcus @trevogre @geoffreyhinton Confabulations is better, but it still sounds too creative and human for what is actually happening :) What was wrong with “mistake”?
Gary Marcus
@GaryMarcus
·
537d
@The_AI_Skeptic @trevogre @geoffreyhinton yes the term is imperfect; i played some role in its popularization, but also tried push for confabulation (which didn’t stick) and welcome other options.
The_AI_Skeptic
@The_AI_Skeptic
·
537d
@trevogre @GaryMarcus @geoffreyhinton Yeah, I dislike the word, too. It’s so misleading. Who coined it and why did it stick, I wonder?
Trevor Green
@trevogre
·
538d
@GaryMarcus @geoffreyhinton LLM's never "hallucinate" because that implies that there is a difference between a wrong answer and a right answer from the LLM, there isn't, it is all the same mechanism. The LLM doesn't know anything, so it's either all hallucinations or none of it is. It is statistical…
Gary Marcus
@GaryMarcus
·
538d
Egregiously misleading rhetoric from @geoffreyhinton.

LLMs can “answer more less any question” but “in a sensible way” is utterly misleading; it abundantly clear that LLMs regularly hallucinate and make boneheaded errors, in essentially every domain that’s been explored.

(Also…

The statement is critical of the term 'mistake' for being insufficient to describe the errors made by AI systems, suggesting that these errors are often not based in reality and are presented as absolute truth. The tone is analytical and seeks to contribute to the discussion on how to properly characterize AI behavior.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not appear to cause harm as it is a critique of terminology rather than an attack on an individual or group. [+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 respects the privacy and dignity of others, as it does not target any individual but rather discusses the conceptual framing of AI errors. [+1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by encouraging a more nuanced discussion of AI behavior. [+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 engages in constructive criticism by suggesting that the term 'mistake' may not be adequate. [+1]
  5. Principle 5:
    I will acknowledge and correct my mistakes.
    The statement acknowledges the complexity of the issue and implies a willingness to correct misunderstandings about AI behavior. [+1]
  6. Principle 6:
    I will use my influence for the betterment of society.
    The statement uses influence to encourage a more accurate understanding of AI systems. [+1]
  7. Principle 7:
    I will uphold the principles of free speech and use my platform responsibly and with integrity.
    The statement upholds the principles of free speech by contributing to a public debate on AI terminology responsibly. [+1]