Gary Marcus

Rank 13 of 47
|
Score 83
Gary Marcus
@GaryMarcus
·
386d
@olcan @mikeknoop @doomie @circlerotator My own thoughts below, referring to this thread. I appreciate people like @doomie speaking up from within the ML community to insist on transparency and proper science.

Think for a minute about @jpineau1’s Replicability Checklist. There is a reason why thrown together…
12/22/2024, 3:50:53 AM
X
In reply to:
Brydon Eastman
@brhydon
·
386d
@GaryMarcus @anupk24 You keep saying 1000 or 750 despite section 1.1 elucidating the true number of training tasks in no uncertain terms.

The literature exists for a reason and opining on an evaluation you're willfully remaining ignorant on is irresponsible scholarship. Full stop.
Gary Marcus
@GaryMarcus
·
386d
@brhydon @anupk24 I am *not* making a stink over the article.

I making a stink over a presentation that many people misinterpreted in part probably because of what Sam said.

I made perfectly clear in other tweets this evening that my concern is not with task, and not with the administration of…
Brydon Eastman
@brhydon
·
386d
@GaryMarcus @anupk24 Please read the paper. It clearly states that the training set is 400 at the top of page 2. You making a stink over an article you've demonstrably not taken the 10 minutes to read is incredibly irresponsible scholarship, especially given the size of your platform.
Gary Marcus
@GaryMarcus
·
386d
@anupk24 @brhydon If Sam had said what you say in the second and third sentences clearly, spelling out what that means concretely (750? training examples), I wouldn’t be objecting.

But I also think significantly fewer people would have come away whispering about AGI.

I look forward to his…
Anup
@anupk24
·
386d
@GaryMarcus @brhydon Sam’s point is that the goal of the model was not to maximize ARC score. The goal was to build a general model that can reason, and part of ARC’s public training set (as allowed by the benchmark) was thrown in, and it achieves the high score.
Gary Marcus
@GaryMarcus
·
386d
@brhydon Singing lessons would be targeting singing.

The statement involves a discussion about the number of training examples used in a machine learning model, referencing a blog and a paper. It critiques the accuracy of information shared by a public figure and emphasizes the importance of responsible scholarship and transparency in scientific communication.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement aims to clarify factual inaccuracies, which aligns with striving to do no harm by promoting accurate information. [+1]
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    The tone is critical but does not engage in personal attacks, respecting the dignity of others involved in the discussion. [+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 by addressing specific inaccuracies and urging responsible scholarship. [+1]