I don't think "judge writing by its quality instead of its source" is coherent. Certainly, the value of reading is in the thoughts it produces in the reader's mind. So if you want me to ignore the source, you're asking me to value the thoughts I get from the writing without regard for the thoughts that produced it. This seems dubious, if not actually oxymoronic. As a reader, it feels much like you telling me to eat something you made, which you enthusiastically assure me is high quality, but which you can't be bothered to eat yourself, and in fact you've never eaten anything like it before. Plus it smells like spray-cheese.
> you're asking me to value the thoughts I get from the writing without regard for the thoughts that produced it. This seems dubious, if not actually oxymoronic
The archaeological record is loaded with texts of dubious or unknown authorship. Many of our oldest classical texts fall into this category [1].
That doesn't make them "oxymoronic" to engage with. Just more difficult. From a time when few were writing and fewer of their works survived, this is worth the effort. When a nickel of electricity can produce text, it's not–we need the filter of authorship to decide how much effort we, as humans, should expend in engaging.
Are you trying to argue that archeological texts weren't written by humans? Because otherwise I don't understand how your comment is related to the parent comment. It has nothing to do with not knowing the identity of the author. It's about the author being a human.
...I'm not sure if you're agreeing with me? The identity of the author is not germane to my point, only whether there actually is an author, aka an actual mind. You could argue those old texts are worth engaging with exactly because they're our only links to those minds.
Ok, but when you're reading Homer or Socrates or the Bible (we're relatively sure it has several different authors), you're still engaging with a communal tradition of fellow readers who, in large part, define what it actually means to understand the text. If you read the Odyssey and decide it was actually about Pokemon, you're misreading it.
> you're still engaging with a communal tradition of fellow readers
Inasmuch as you’re reading translations upon translations, sure. But none of that is necessary.
> If you read the Odyssey and decide it was actually about Pokemon, you're misreading it
If you can stand by your reading based on the text, I say go for it. It doesn’t work in your example. But there are plenty of cases where what was once an against-the-grain interpretation dived into and out of the mainstream.
Many people here complain that AI writing is bad. And I agree with it, AI writing is boring and its distinctive tone is now associated with unpleasant thoughts for most people including me.
But I'm curious about how people will feel when AI writing is good and impossible to differentiate from a decent journalist-level writer, including with tooling. Then what? Are people going to stop reading?
I'm trying to separate the part of the argument related to "this writing is bad" from "this was written by AI".
I personally don't care if AI is a writing instrument for a human who's signing off on the content, as long as the writing is good and the source is legitimate.
The fundamental problem with most AI "writing" is that writing is a communication medium, and if your text was just generated by a prompt, I would always rather read the prompt than the output, as the output contains zero additional bits of information.
I don't want to see the peer review process (which can be weak enough as it is) overwhelmed by a flood of slop. I don't want reviewers to have to suffer reading AI's bad writing either.
"Because I was first trained as a mathematician, I probably write more like an AI than most people. But that particular article had been through 10 drafts. I worked on it with Claude Fable 5.1 and other large language models, as I now work on nearly everything, but I reviewed every line and rewrote most of them. By draft 10, I could not have said where the machine’s contribution stopped and mine began.?"
...
"The author drafted this essay in extended collaboration with Claude Fable 5.1."
-----------------
If you read a lot of journal articles, you're probably accustomed to occasionally running into articles a ways outside your bailiwick that are written poorly, you don't really understand, which you suspect made it past reviewers who didn't really understand it, and indeed, you suspect the author didn't really understand it either. Everyone involved just gave them the benefit of the doubt because they were too busy to figure out what the tortured language really meant. However, you don't know for sure, so you decide to leave it up to people in that particular field to deal with.
I can't speak for others, but I find AI writing to be frequently similar in its effect on me to bad writing. Yes, the grammar is correct, but it beats around the bush and puts emphasis on the wrong things. Even if you filter Claude through a human, those poor choices tend to shine through. This subtly inhuman voice switches your brain off. I suspect a lot of bad science is going to slip past reviewers because of LLM writing, and a lot of good science is going to slip past those who might need to read it, because the LLM co-written article switched their brains off and failed to make itself understood.
I understand the notion that, if the content is good enough, it shouldn't matter if the article is written poorly or co-written with AI. However, if the content is good enough, isn't it worth communicating it well?
Good point about emphasis. Correctness in emphasis, and I do mean correctness, was already underrated. It's a critical measure of whether you actually understand a topic, and an indicator of values in general.
If AI wrote it, it's not your voice. If you can't communicate your thoughts then it reflects a lack of clarity on you as the author. If you lack the clarity, AI assisted anything is probably part of the problem.
If connecting with the audience and guiding them to understand your thoughts is irrelevant, then you are missing the point of writing and being an academic.
There is a reason AI slop is revolting to read. It doesn't care.
I am on two program committees for major CS conferences this year.
LLM writing is horrible, full stop. The vast majority of papers that I see with obvious LLM writing ticks are also frankly incomprehesible. Claude's biggest sin at the moment is inventing new terminology (sometimes multiple times in a single sentence!) without stopping to explain what it means.
In previous years, I would see a substantial fraction (maybe 30-50%) of papers that were obviously written by non-native English speakers, and full of grammatical mistakes and typos. I have seen zero papers like that this year. Honestly, I preferred those: reading LLM writing is painful by comparison.
So far I have seen exactly one paper that had obvious LLM writing ticks, and yet had gone through enough passes (either of the LLM or the human authors) that the paper itself was comprehensible. So I'll admit that it's possible to do. But based on what I've seen this is vastly in the minority.
I honestly used to enjoy the review process; even when the papers were bad at least I could usually see the effort that went into it and give some constructive guidance on how to improve. The process is dramatically less pleasant now, and it's not obvious if I'm accomplishing anything: if authors who write with LLMs just stick the reviews back into those same LLMs, will they learn anything? I suspect not.
Maybe some day LLMs will be good enough to write, but as of today: please, please, just don't do it. The typos and grammatical mistakes are easier to read than the slop.
To be honest, people overvalue human writing too much. Even in academic research, there were quite a lot of cases where the writing was strange. I actually think AI writing is more readable, to that extent. People sometimes compare AI writing with the very best of human writing, but considering that AI writing is above average, I think it’s readable enough.
I'm fine with using AI for code, where the quality will likely be better than those of academics.
But writing.. FFS everything I see is just AI slop. And I only realize when I start wondering why the paper is going and going and saying nothing, and then pangram tells me that everything past the intro is just AI. :/
The glass half full take is that we are very lucky that there is some much literature that pre-dates LLMs. Our brains are tuned to enjoy that more. People who grow up with LLM slop from the start won't necessarily see it the same way.
Yup. I can read novels from the 1st century [1] or from the 19th century [2] or from the 1970s [3] for the rest of my life (well, until they get banned).
I care more if work is reproducible. Most academics articles are "trust me bro, I am scientist", no source data, no scripts used to process them. With AI I would expect publication of prompts and harness settings.
I don't think "judge writing by its quality instead of its source" is coherent. Certainly, the value of reading is in the thoughts it produces in the reader's mind. So if you want me to ignore the source, you're asking me to value the thoughts I get from the writing without regard for the thoughts that produced it. This seems dubious, if not actually oxymoronic. As a reader, it feels much like you telling me to eat something you made, which you enthusiastically assure me is high quality, but which you can't be bothered to eat yourself, and in fact you've never eaten anything like it before. Plus it smells like spray-cheese.
> you're asking me to value the thoughts I get from the writing without regard for the thoughts that produced it. This seems dubious, if not actually oxymoronic
The archaeological record is loaded with texts of dubious or unknown authorship. Many of our oldest classical texts fall into this category [1].
That doesn't make them "oxymoronic" to engage with. Just more difficult. From a time when few were writing and fewer of their works survived, this is worth the effort. When a nickel of electricity can produce text, it's not–we need the filter of authorship to decide how much effort we, as humans, should expend in engaging.
[1] https://en.wikipedia.org/wiki/Homer#Biography,_identity,_and...
Are you trying to argue that archeological texts weren't written by humans? Because otherwise I don't understand how your comment is related to the parent comment. It has nothing to do with not knowing the identity of the author. It's about the author being a human.
...I'm not sure if you're agreeing with me? The identity of the author is not germane to my point, only whether there actually is an author, aka an actual mind. You could argue those old texts are worth engaging with exactly because they're our only links to those minds.
> whether there actually is an author, aka an actual mind
The mind is long dead. In Cartesian form, we assume it existed. But it may not be, as with Homer, a single mind. We know it didn’t even think like us.
One can engage with text on its own. Even find meaning in it. Being unable and refusing to are separate categories.
Ok, but when you're reading Homer or Socrates or the Bible (we're relatively sure it has several different authors), you're still engaging with a communal tradition of fellow readers who, in large part, define what it actually means to understand the text. If you read the Odyssey and decide it was actually about Pokemon, you're misreading it.
> you're still engaging with a communal tradition of fellow readers
Inasmuch as you’re reading translations upon translations, sure. But none of that is necessary.
> If you read the Odyssey and decide it was actually about Pokemon, you're misreading it
If you can stand by your reading based on the text, I say go for it. It doesn’t work in your example. But there are plenty of cases where what was once an against-the-grain interpretation dived into and out of the mainstream.
Many people here complain that AI writing is bad. And I agree with it, AI writing is boring and its distinctive tone is now associated with unpleasant thoughts for most people including me.
But I'm curious about how people will feel when AI writing is good and impossible to differentiate from a decent journalist-level writer, including with tooling. Then what? Are people going to stop reading?
I'm trying to separate the part of the argument related to "this writing is bad" from "this was written by AI".
I personally don't care if AI is a writing instrument for a human who's signing off on the content, as long as the writing is good and the source is legitimate.
The fundamental problem with most AI "writing" is that writing is a communication medium, and if your text was just generated by a prompt, I would always rather read the prompt than the output, as the output contains zero additional bits of information.
I care. I don't want to see the brute forcing of academic publication
I don't want to see the peer review process (which can be weak enough as it is) overwhelmed by a flood of slop. I don't want reviewers to have to suffer reading AI's bad writing either.
https://archive.is/BrQB7
"Because I was first trained as a mathematician, I probably write more like an AI than most people. But that particular article had been through 10 drafts. I worked on it with Claude Fable 5.1 and other large language models, as I now work on nearly everything, but I reviewed every line and rewrote most of them. By draft 10, I could not have said where the machine’s contribution stopped and mine began.?"
...
"The author drafted this essay in extended collaboration with Claude Fable 5.1."
-----------------
If you read a lot of journal articles, you're probably accustomed to occasionally running into articles a ways outside your bailiwick that are written poorly, you don't really understand, which you suspect made it past reviewers who didn't really understand it, and indeed, you suspect the author didn't really understand it either. Everyone involved just gave them the benefit of the doubt because they were too busy to figure out what the tortured language really meant. However, you don't know for sure, so you decide to leave it up to people in that particular field to deal with.
I can't speak for others, but I find AI writing to be frequently similar in its effect on me to bad writing. Yes, the grammar is correct, but it beats around the bush and puts emphasis on the wrong things. Even if you filter Claude through a human, those poor choices tend to shine through. This subtly inhuman voice switches your brain off. I suspect a lot of bad science is going to slip past reviewers because of LLM writing, and a lot of good science is going to slip past those who might need to read it, because the LLM co-written article switched their brains off and failed to make itself understood.
I understand the notion that, if the content is good enough, it shouldn't matter if the article is written poorly or co-written with AI. However, if the content is good enough, isn't it worth communicating it well?
Good point about emphasis. Correctness in emphasis, and I do mean correctness, was already underrated. It's a critical measure of whether you actually understand a topic, and an indicator of values in general.
Lets go further: Who cares wheter biological humanoids use ai to think?
Aside from the people who have to live with the results of outsourced thinking, most notably that would be the people selling the thinking.
If AI wrote it, it's not your voice. If you can't communicate your thoughts then it reflects a lack of clarity on you as the author. If you lack the clarity, AI assisted anything is probably part of the problem.
If connecting with the audience and guiding them to understand your thoughts is irrelevant, then you are missing the point of writing and being an academic.
There is a reason AI slop is revolting to read. It doesn't care.
> Allison Stanger is a professor of political science at Middlebury College.
Is that so? And that's about it?
I am on two program committees for major CS conferences this year.
LLM writing is horrible, full stop. The vast majority of papers that I see with obvious LLM writing ticks are also frankly incomprehesible. Claude's biggest sin at the moment is inventing new terminology (sometimes multiple times in a single sentence!) without stopping to explain what it means.
In previous years, I would see a substantial fraction (maybe 30-50%) of papers that were obviously written by non-native English speakers, and full of grammatical mistakes and typos. I have seen zero papers like that this year. Honestly, I preferred those: reading LLM writing is painful by comparison.
So far I have seen exactly one paper that had obvious LLM writing ticks, and yet had gone through enough passes (either of the LLM or the human authors) that the paper itself was comprehensible. So I'll admit that it's possible to do. But based on what I've seen this is vastly in the minority.
I honestly used to enjoy the review process; even when the papers were bad at least I could usually see the effort that went into it and give some constructive guidance on how to improve. The process is dramatically less pleasant now, and it's not obvious if I'm accomplishing anything: if authors who write with LLMs just stick the reviews back into those same LLMs, will they learn anything? I suspect not.
Maybe some day LLMs will be good enough to write, but as of today: please, please, just don't do it. The typos and grammatical mistakes are easier to read than the slop.
To be honest, people overvalue human writing too much. Even in academic research, there were quite a lot of cases where the writing was strange. I actually think AI writing is more readable, to that extent. People sometimes compare AI writing with the very best of human writing, but considering that AI writing is above average, I think it’s readable enough.
I'm fine with using AI for code, where the quality will likely be better than those of academics.
But writing.. FFS everything I see is just AI slop. And I only realize when I start wondering why the paper is going and going and saying nothing, and then pangram tells me that everything past the intro is just AI. :/
The glass half full take is that we are very lucky that there is some much literature that pre-dates LLMs. Our brains are tuned to enjoy that more. People who grow up with LLM slop from the start won't necessarily see it the same way.
Yup. I can read novels from the 1st century [1] or from the 19th century [2] or from the 1970s [3] for the rest of my life (well, until they get banned).
[1] https://en.wikipedia.org/wiki/Callirhoe_(novel)
[2] https://en.wikipedia.org/wiki/Jane_Eyre
[3] https://en.wikipedia.org/wiki/Swords_and_Deviltry
I care more if work is reproducible. Most academics articles are "trust me bro, I am scientist", no source data, no scripts used to process them. With AI I would expect publication of prompts and harness settings.