Post

I can elaborate on why I think it's a bad idea (starting with the opt-out v. opt-in aspect). But it boils down to the fact that so far, all improvements I have seen thanks to the AI tools are papers that are smoother, but also less informative. Fewer typos and grammatical mistakes, but less content.

1 repost25 likes

So in reality you're mostly annoyed because before you could use typos and grammatical errors as a proxy to detect weak papers due to some correlation between the two, whereas now it requires reading them carefully?

We're doing a complete disservice to everyone by pretending that the style (fewer typos and better use of a thesaurus) matters more than the substance. I don't read a paper to have paragraphs and paragraphs of polished platitudes.

Of course. But on the other hand, language does matter. As an example, I'm reviewing a letters paper at the moment and my review is essentially going to be "guys, the English in your paper is so broken that I cannot decipher what you're doing. Please rewrite everything clearly and resubmit".

No matter how you think about this, I do not think Overleaf is to blame. They are not enabling a new capability. They making an existing capability (AI text polishing), that was heavily used already, easier to use. If they would not do it, an competitor might emerge that does.

Yep, let's blame the game - and I have mixed feelings about this myself. Regarding Overleaf: They are a company providing a text editor, not the guardians of the ivory tower. They provide the services they think most their users want (including opt-in vs opt-out).

Yes, AI writing does not have sufficient precision for scientific writing. The devil's in the details and AI just doesn't get them right in my experience. There's still no replacement for careful deliberation in the choice of words