Gary Marcus

Rank 23 of 47
|
Score 47
9/13/2024, 5:57:31 PM
X
In reply to:
Julian
@mealreplacer
·
341d
@GaryMarcus Are you saying that they are deliberately fine-tuning the models on these examples? Why?
Gary Marcus
@GaryMarcus
·
341d
@mealreplacer Yeah ‘cause they train on them. And then when you chance the problem trivially they often still fail, eg the strawberry letter thing, updated for a case less popular:
Gary Marcus
@GaryMarcus
·
341d
OpenAI has zero t’s left to give.
Julian
@mealreplacer
·
342d
(TBC: I don't think this is huge evidence of the model's capabilities, or whatever. It's just funny to see whether N+1 models can solve whatever funny little blunders Gary cherry picks for any given N)
Julian
@mealreplacer
·
342d
Gary Marcus
@GaryMarcus
·
462d
Someday my children will ask me why anyone ever took seriously the thought that systems like this had anything to do with AGI.

I honestly have no idea how I will answer. Mass delusion?

(Example emailed to me by a friend.)

The statement 'Yes. I have been told this, and they often do much better on exact formulations than on minor variations.' is part of a broader discussion about the capabilities and limitations of AI models, specifically in the context of fine-tuning and problem-solving. The tone is informative and seems to be aimed at clarifying a point about AI behavior in response to a query.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does no harm and is neutral in tone, focusing on providing information rather than criticism or negative commentary. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by providing a clear explanation about the behavior of AI models, which can help others better grasp the nuances of AI training and performance. [+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 dialogue by addressing a specific question and providing a thoughtful response without resorting to personal attacks or ad hominem arguments. [+1]