Gary Marcus

Rank 14 of 47
|
Score 89
3/15/2024, 10:51:27 PM
X
In reply to:
Derya Unutmaz, MD
@DeryaTR_
·
570d
@GaryMarcus @linasvepstas @say_cem In real world? Ok Here is an experiment to try : cross check what GPT4 produces with Claude 3 and Gemini Pro. Given different training sets, I suspect they are better to fact check if there are ground truths.
Gary Marcus
@GaryMarcus
·
570d
@DeryaTR_ @linasvepstas @say_cem In the real world it didn’t
Derya Unutmaz, MD
@DeryaTR_
·
570d
@linasvepstas @GaryMarcus @say_cem LLMs also offer potential way to cross check those flaws. As pointed out by @say_cem an LLM would have picked up this but human editor didn’t.
Linas
@linasvepstas
·
570d
@DeryaTR_ @GaryMarcus @say_cem Human nature is flawed. We have created political and social systems that attempt to limit the damage caused by those flaws. LLM's offer brand-new ways of making end-runs around those regulatory systems. Political and social systems need to adjust. But how?
Derya Unutmaz, MD
@DeryaTR_
·
570d
@GaryMarcus @linasvepstas @say_cem I am not pretending it’s fine. My point is that this has nothing to do with LLMs. Most papers are of poor quality, and science advances because of a minority of publications; this has always been the case.

So please stop blaming the LLM for the flaws in human nature.
Gary Marcus
@GaryMarcus
·
570d
@DeryaTR_ @linasvepstas @say_cem Sure. But given how atrociously lax the editing was by both the reviewers and the authors on the introduction, you can’t trust that the study or its discussion or analysis was carefully reviewed, nor that that is was written by humans.

Stop pretending like this is totally fine.

The statement expresses concern about the practical challenges of using language learning models (LLMs) in academic publishing, highlighting the issue of errors slipping past busy authors and reviewers. The tone is serious and addresses a real-world problem in scholarly communication.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not intend to harm but rather to point out a concern in the academic process. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement lacks a promotion of understanding, empathy, or compassion, as it does not offer solutions but focuses on the problem. [-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 identifying a problem but does not engage in dialogue to resolve disagreement.