I frequently am accused of using LLMs to write my prose, something that I not only eschew, but also believe is morally corrupt and intellectually dishonest.
I’m not above spellcheck, grammar checkers, or even LLM driven evaluation of articles, but my thoughts, word choices, and structure are always of my own design.
I use the em-dash where it is appropriate.
I find that people accusing writers of using AI typically disagree with the premise of the text, and use the “AI” character assault as a method of dehumanising the author and dismissal of their work. The assertion is very rarely made in good faith, but rather is used as a weak attempt to discredit an idea without actually refuting the premise or even examining the argument.
Shame on whoever argues in this way, it’s weak, unproductive, and intellectually lazy. It’s fine to disagree, but if you aren’t willing to act in good faith, just keep your thoughts to yourself. You’re only going to discredit your own point of view if you touch the keyboard.
I use the em dash as appropriate, similar to semicolons and their ilk.
I don't think use of an em dash is indicative in itself of AI assistance, but rather, the change to using them. Did this person all of a sudden start using them? There are also other things to look at, like how certain bullet point lists have emphasis (for key phrases, being bold, when previously the author didn't do so, stylistically).
I write a lot (as a PM) - I've taken to using MacWhisper, which does local AI dictation, but also (at my configuration) sends it to a ChatGPT prompt first:
"You are a professional proofreader and editor. Your task is to refine and polish the given transcript as follows:
1. Correct any spelling errors.
2. Fix grammatical mistakes.
3. Improve punctuation where necessary.
4. Ensure consistent formatting.
5. Clarify ambiguous phrasing without changing the meaning.
6. If a sentence or paragraph is overly verbose and has more than negligible redundancy, lightly edit for brevity.
7. If the transcript contains a question, edit it for clarity but do not provide an answer.
Please return only the cleaned-up version of the transcript. Do not add any explanations or comments about your edits."
This is great. I get the benefits of pretty accurate transcription while getting a first pass at copyediting almost in real time. It did require me to make some tweaks to my dictation process (allowing it to "chew" on larger chunks to give better context to its editing), but it works very well.
Long live the em-dash!
I frequently am accused of using LLMs to write my prose, something that I not only eschew, but also believe is morally corrupt and intellectually dishonest.
I’m not above spellcheck, grammar checkers, or even LLM driven evaluation of articles, but my thoughts, word choices, and structure are always of my own design.
I use the em-dash where it is appropriate.
I find that people accusing writers of using AI typically disagree with the premise of the text, and use the “AI” character assault as a method of dehumanising the author and dismissal of their work. The assertion is very rarely made in good faith, but rather is used as a weak attempt to discredit an idea without actually refuting the premise or even examining the argument.
Shame on whoever argues in this way, it’s weak, unproductive, and intellectually lazy. It’s fine to disagree, but if you aren’t willing to act in good faith, just keep your thoughts to yourself. You’re only going to discredit your own point of view if you touch the keyboard.
I use the em dash as appropriate, similar to semicolons and their ilk.
I don't think use of an em dash is indicative in itself of AI assistance, but rather, the change to using them. Did this person all of a sudden start using them? There are also other things to look at, like how certain bullet point lists have emphasis (for key phrases, being bold, when previously the author didn't do so, stylistically).
I write a lot (as a PM) - I've taken to using MacWhisper, which does local AI dictation, but also (at my configuration) sends it to a ChatGPT prompt first:
"You are a professional proofreader and editor. Your task is to refine and polish the given transcript as follows:
1. Correct any spelling errors.
2. Fix grammatical mistakes.
3. Improve punctuation where necessary.
4. Ensure consistent formatting.
5. Clarify ambiguous phrasing without changing the meaning.
6. If a sentence or paragraph is overly verbose and has more than negligible redundancy, lightly edit for brevity.
7. If the transcript contains a question, edit it for clarity but do not provide an answer.
Please return only the cleaned-up version of the transcript. Do not add any explanations or comments about your edits."
This is great. I get the benefits of pretty accurate transcription while getting a first pass at copyediting almost in real time. It did require me to make some tweaks to my dictation process (allowing it to "chew" on larger chunks to give better context to its editing), but it works very well.