Gary Marcus

Rank 14 of 47
|
Score 92
4/8/2024, 10:10:34 PM
X
In reply to:
Kevin Kohler
@KevinKohlerFM
·
536d
@GaryMarcus @patmcguinness @geoffreyhinton I have read it & appreciate your work. But wouldn't it be fair to say the bar for "generalization beyond training space" has changed a bit? LLM's have no problem to deal with a made-up subject in a sentence, or applying grammatical rules to a made-up verb. 1998 excerpt attached
Gary Marcus
@GaryMarcus
·
536d
@patmcguinness @geoffreyhinton Read my 1998 paper and then the paper on exponential data yesterday. They are essentially the same argument. Goalposts have not moved.
Dr Patrick M. - AI Builder
@patmcguinness
·
536d
@GaryMarcus @geoffreyhinton "LLMs regularly hallucinate and make boneheaded errors, in essentially every domain that’s been explored."

True, they do, but so do people. Imperfections & limits don't negate Hinton's point that these are general systems, ie "AGI" but we moved goalposts on that term.
Gary Marcus
@GaryMarcus
·
536d
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 part of a technical discussion on AI, referencing academic research to support a claim about machine learning model training spaces.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and is focused on sharing academic findings. [+1]
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    There is no violation of privacy or dignity in the statement. [+1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement contributes to understanding by referencing relevant research. [+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 is part of a constructive dialogue on AI model transparency and performance. [+1]
  5. Principle 5:
    I will acknowledge and correct my mistakes.
    There is no indication of a mistake being made or corrected.
  6. Principle 6:
    I will use my influence for the betterment of society.
    The sharing of academic research can be seen as using influence to better society by promoting informed discussions on AI. [+1]
  7. Principle 7:
    I will uphold the principles of free speech and use my platform responsibly and with integrity.
    The statement upholds free speech and uses the platform responsibly by engaging in a substantive technical debate. [+1]