Having read your subsequent article where you give us both papers to read (it's pretty obvious which is which), of course it's no surprise that the LLMs like the LLM-written version (what else would you expect?) I notice that everything in the LLM version is pared down to the minimum, almost as if the whole paper is a succession of tweets. Maybe that sort of paper is more accessible to short-format readers (who seem to be on the increase these days) but don't you think the medium itself (of the academic paper) deserves some respect in its own right? I like what Dawn Song did with her peer preservation paper at U Berkeley: she releases a blog with the paper. The paper is properly academic. The blog (where I think the LLM style has more of a point) is focused on key messaging to a broader audience. https://rdi.berkeley.edu/blog/peer-preservation/
I can't help but have a viscerally negative reaction to your self-doubt about whether your academic paper is better than the AI written one. I read what you write not only because it is insightful, but because it is you who have written it. I can tolerate a bit of post-writing editing from an AI model, but if the writing itself ends up not being your work, I'm not necessarily interested in reading it. I imagine many of your academic peers would feel similarly. Please don't forget that writing itself is a connection between persons: words from your mind make your mind known to mine. Please do not undervalue this.
Thanks John! And this is very much where I am coming from, and to be honest finding it disheartening seeing how many people are beginning to side with AI prose. There also seems to be a feedback loop here where LLMs are learning favor LLM-style writing above what is authentically human ...
Having read your subsequent article where you give us both papers to read (it's pretty obvious which is which), of course it's no surprise that the LLMs like the LLM-written version (what else would you expect?) I notice that everything in the LLM version is pared down to the minimum, almost as if the whole paper is a succession of tweets. Maybe that sort of paper is more accessible to short-format readers (who seem to be on the increase these days) but don't you think the medium itself (of the academic paper) deserves some respect in its own right? I like what Dawn Song did with her peer preservation paper at U Berkeley: she releases a blog with the paper. The paper is properly academic. The blog (where I think the LLM style has more of a point) is focused on key messaging to a broader audience. https://rdi.berkeley.edu/blog/peer-preservation/
I can't help but have a viscerally negative reaction to your self-doubt about whether your academic paper is better than the AI written one. I read what you write not only because it is insightful, but because it is you who have written it. I can tolerate a bit of post-writing editing from an AI model, but if the writing itself ends up not being your work, I'm not necessarily interested in reading it. I imagine many of your academic peers would feel similarly. Please don't forget that writing itself is a connection between persons: words from your mind make your mind known to mine. Please do not undervalue this.
Thanks John! And this is very much where I am coming from, and to be honest finding it disheartening seeing how many people are beginning to side with AI prose. There also seems to be a feedback loop here where LLMs are learning favor LLM-style writing above what is authentically human ...
Much more on this in my follow-on post.