There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe nicely stated
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe There is a subtle difference between things that are well-formed and things that are correct/good. Obviously it's possible for something to be both at once, or neither. But one doesn't perfectly imply the other.
This was probably the heart of the aggrieved feeling I got back when I was trying to write a lot of Rust code. The borrow checker tends to push you in a direction of well-formed trash. You eventually get so cognitively exhausted that you just surrender to it. I profoundly disliked that feeling of architectural-scope brain fog.
-
@knack for my technical documentation covering software mostly written by AI, I find I have to start with Claude-written docs. But, so far, my process has been to just read the docs and try to follow them, and fix where it's hard, while also cleaning up all the weird tics and verbosity. Often, by the time I'm finished with them, there's very little AI written prose left, because I find it so off-putting and hard to follow (though Opus 5.5 has improved somewhat over the incoherent Opus 5).
-
@knack for my technical documentation covering software mostly written by AI, I find I have to start with Claude-written docs. But, so far, my process has been to just read the docs and try to follow them, and fix where it's hard, while also cleaning up all the weird tics and verbosity. Often, by the time I'm finished with them, there's very little AI written prose left, because I find it so off-putting and hard to follow (though Opus 5.5 has improved somewhat over the incoherent Opus 5).
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe If there is no point, the author may not have given it one. I have had success with Chat GPT writing a relatively short piece with a very specific prompt. Essentially you give it an outline and tell it who your audience is. Once you have a base version, you can try out variations.
-
@swelljoe If there is no point, the author may not have given it one. I have had success with Chat GPT writing a relatively short piece with a very specific prompt. Essentially you give it an outline and tell it who your audience is. Once you have a base version, you can try out variations.
I wouldn't use AI if I wanted the writing to sound like me, but if you just need something generic, it's fine.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe Thanks for defining this! I've been heading in the same direction.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe I agree, I absolutely despise AI prose, and I argue hard for each client to push AI to generate and mutate UI components to make a point, rather than just spit out prose, works way better
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe This might just be me, but no English teacher worth their salt has ever cared about how grammatically defensible a paper is. Sure they teach the rules of Standard English and highlight when writing doesn't conform with Standard English, but the focus is generally on improving the clarity of the student's message.
But yes, people who write in non-Standard English offer amazing insights! -
@swelljoe This might just be me, but no English teacher worth their salt has ever cared about how grammatically defensible a paper is. Sure they teach the rules of Standard English and highlight when writing doesn't conform with Standard English, but the focus is generally on improving the clarity of the student's message.
But yes, people who write in non-Standard English offer amazing insights!@bamfingle also true. Just saying pedantic adherence to grammar rules isn't necessarily improving the effectiveness of ones writing, and LLMs, while good at the rules, are bad at a more important dimension of what makes for quality writing, and I think relying heavily on them for editing probably makes ones writing worse overall, even if it looks more proper. That's anecdotal, and I know a lot of people are relying on it for editing, and, worse, writing.
But, I'm going to stick to writing all my own posts (blog, social, etc.) and emails, and for technical docs about software written by LLMs, I'm going to edit the hell out of it.
-
@bamfingle also true. Just saying pedantic adherence to grammar rules isn't necessarily improving the effectiveness of ones writing, and LLMs, while good at the rules, are bad at a more important dimension of what makes for quality writing, and I think relying heavily on them for editing probably makes ones writing worse overall, even if it looks more proper. That's anecdotal, and I know a lot of people are relying on it for editing, and, worse, writing.
But, I'm going to stick to writing all my own posts (blog, social, etc.) and emails, and for technical docs about software written by LLMs, I'm going to edit the hell out of it.
@bamfingle I've also seen teachers basically pleading with their students to not rely on LLMs, because they'd rather see interesting and original thoughts than perfect English.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe The biggest problem with “AI” is that there is - nor can there ever be - “I”. It is an insightless roulette wheel which vomits word salad.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
@swelljoe Nicely put. I've been feeling the same, there's definitely a sort of squidgy unsavoury feeling I get when reading LLM prose but I couldn't really put my finger on what it was until now.
-
There's an interesting thing I've noticed about myself with the onslaught of hard-to-read AI prose seemingly everywhere. I am historically a rules lawyer (in my head, not out loud) about grammar and spelling, silently judging folks for ugly or clunky prose. AI writing has softened my position on this issue. LLMs write with excellent grammar, but they have no intention, so they can't ever really get to the point. They don't want anything, so there is no point to anything they say. They are generally bad writers because of it.
My fixation when reading now is on whether there is a clear point, something the author wants me to get, not whether the grammar is nice or the spelling is good.
I think this is telling me that I, and you, shouldn't use AI even to edit and definitely not to write. I suspect it makes your writing worse even while making it more grammatically defensible to an English teacher.
I completely agree with you when it comes to creative writing. On the other hand, AI systems and LLMs are an incredibly powerful tool that I wouldn’t want to do without anymore, because they can speed up work tremendously and actually improve it in several areas. For example, I dictated this reply and had an AI convert it into text. Afterwards, an LLM translated it into English.
1/2
-
I completely agree with you when it comes to creative writing. On the other hand, AI systems and LLMs are an incredibly powerful tool that I wouldn’t want to do without anymore, because they can speed up work tremendously and actually improve it in several areas. For example, I dictated this reply and had an AI convert it into text. Afterwards, an LLM translated it into English.
1/2
To stick with your example: this has enabled me to easily put my thoughts down on paper and convert them into the language we both speak. I think that’s great.
2/2
-
To stick with your example: this has enabled me to easily put my thoughts down on paper and convert them into the language we both speak. I think that’s great.
2/2
@SaltyMonk yeah, I'm not actually broadly anti-AI, though I think a lot of folks are taking it that way. It's incredible for accessibility, translation, and, yes, writing code. I'm ambivalent about that last one, but it's a quite large lever for tackling projects I couldn't have previously tackled solo (with any realistic completion date).
I'm specifically talking about prose, a thing that really shows what LLMs are in a way that I don't think is obvious in most pursuits. They've been trained to write like people, but it's surface-level stuff. The lack of intention/desire/hopes/fears/memories makes their writing a fundamentally different thing. Humans bob and weave through all their competing urges and feelings and thoughts and put it down on paper and the result is more interesting for it. And, so far, that's a thing that cannot be effectively simulated with a mere prompt.
-
@SaltyMonk yeah, I'm not actually broadly anti-AI, though I think a lot of folks are taking it that way. It's incredible for accessibility, translation, and, yes, writing code. I'm ambivalent about that last one, but it's a quite large lever for tackling projects I couldn't have previously tackled solo (with any realistic completion date).
I'm specifically talking about prose, a thing that really shows what LLMs are in a way that I don't think is obvious in most pursuits. They've been trained to write like people, but it's surface-level stuff. The lack of intention/desire/hopes/fears/memories makes their writing a fundamentally different thing. Humans bob and weave through all their competing urges and feelings and thoughts and put it down on paper and the result is more interesting for it. And, so far, that's a thing that cannot be effectively simulated with a mere prompt.
@swelljoe Full ack!
-
J jwcph@helvede.net shared this topic
