There is only one way to eliminate AI writing from your reading list and that is to accept nothing less than truly inspired, original writing. I no longer look for "tells"or "giveaways". If it feels stale I lose Interest. The goal is not to weed out botwork but to read good stuff. If AI inspired me who cares what wrote it? I enjoy a bit of quirky wordplay, not enough to slow down the read but it's okay for the writer to show every now and then. Nobody likes their lily so guilded that you can't see the pond for the pads but it's better than that too-many-cheetos feeling you get from reading a procession of pretentious plebian pedantries plucked out of last weeks Medium staff picks.
I'm working towards my doctorate right now and have been plugging away on my dissertation. My chair advised me to be sure to remove all AI from my work. There is no AI in my work. However, when I upload to Grammarly, it will invariably report a percentage of AI "patterns" in my work.
So I played a little trick on AI and Grammarly. I had AI write a one-page essay on something and uploaded it to Grammarly for a quick check - only 4% AI patterns recognized.
You are right. There is no detection system available, and as soon as someone invents one, the AI bots will invent a tool to avoid detection.
You nailed it. I had a submission rejected last month from a popular online publication I hadn't worked with before--they said they don't accept AI written material. Don't know what detector they used, but it failed 100%
That is so frustrating. I'm sure that there will eventually be lawsuits against academic institutions (and now publications) that accuse students and writers of using AI.
If you don't mind, I'll offer some perspective from a software engineer's standpoint, since I was doing that for years, too. Part of the reason current LLM models so frequently generate the kind of stuff you refer to is that current models are really just huge databases (really, really huge) of prior art.
AI processes are even distantly similar to some of the existing search tools that software people have been using for years, like SOLR, which derives its "intelligence" by matching patterns around letters so that when someone searches for music but types, "mustic" instead, it returns "music"-related stuff, because SOLR is aware of established character patterns in that word.
So AI is really just a big, fat old pattern-matching engine.
Their pattern matching is of such enormous complexity that there is controversy over the enormous amounts of energy they consume. It's too bad people insist on using it, because these things require massive data centers that directly conflict with goals for addressing climate change, but that's a different comment and post, I guess.
Yes, a different post--and I wish you would write it. Not many people know the toll this could take carbon-footprint-wise. Thanks for this insight--very interesting and completely worthwhile!
Thanks! Yeah, I have a few AI-related articles in my various queues here and on Medium. I haven't had a chance to flesh them out. I've run one or two, as well. Linda Carroll just posted an excellent one today.
I'd like to know your opinion on using AI as an editor for spell check, grammar, and punctuation. I don't like Grammarly, never have. I have issues with using passive voice and that's mostly what I use it for. I don't know if using AI is good or bad for this, but I do use it for something other than writing, so pay for a subscription.
The one thing I notice when using AI when it comes to punctuation. It uses a lot of em dashes. I use a lot of them too, I like dashes and ellipsis, I write like I think. It's me and I'm not changing that, but noticed it likes to add them into places I wouldn't typically use them.
And for the record... my dad is rolling in his grave right now knowing I can't use the word 'plethora' - it was a word inside a family joke, used for years, and it makes me smile when I see it.
Loved this, Maryan. It was interesting to reading gives me things to think about.
Thank you, Patti - you're always so supportive. I write the way write, growing and polishing as time continues to go by. As for AI for grammar, I have used Gemini and GPT, but I prefer Grammarly to do it inline as I write--saves me time. I ignore almost everything Grammarly tells me except for actual rules of grammar stuff. I don't need a bot to refine my style. I set the settings in Grammarly to follow my rules. If you hate that one, you might try Quillbot--
I didn't know I could set the settings in Grammarly. I'll go back and revisit, maybe learning to use it might help. I'll check out Quillbot, thank you for the suggestion.
Elevate, unlock, and revolutionize are among many that are overused by AI and I've banned it from using in my custom instructions and memory (ChatGPT).
I find that AI writing sounds voiceless and generic. It reads like the instructions for opening a cereal box even as it’s allegedly about a personal experience. It reads without any variety in the sentence length. That’s a major AI tell over buzzwords that humans will use with impunity so long as someone who is winning at life is using them.
But equally, I don’t care if someone thinks what I write is AI because anyone who doesn’t trust themselves isn’t going to trust me and that says a hell of a lot more about them than me, so by all means, the tools are always going to assume that other people are tools and are using tools to cheat and squeak by just like they would. And those people give me ewe.
Haha, this is funny. I never thought about that. I like AI, but writing comes from my heart and it isn't going to take away my DNA thing for my doing so... if people don't want to read me, that's okay, too. The trust issue is interesting, glad you brought it up. Thank you.
There is only one way to eliminate AI writing from your reading list and that is to accept nothing less than truly inspired, original writing. I no longer look for "tells"or "giveaways". If it feels stale I lose Interest. The goal is not to weed out botwork but to read good stuff. If AI inspired me who cares what wrote it? I enjoy a bit of quirky wordplay, not enough to slow down the read but it's okay for the writer to show every now and then. Nobody likes their lily so guilded that you can't see the pond for the pads but it's better than that too-many-cheetos feeling you get from reading a procession of pretentious plebian pedantries plucked out of last weeks Medium staff picks.
I'm working towards my doctorate right now and have been plugging away on my dissertation. My chair advised me to be sure to remove all AI from my work. There is no AI in my work. However, when I upload to Grammarly, it will invariably report a percentage of AI "patterns" in my work.
So I played a little trick on AI and Grammarly. I had AI write a one-page essay on something and uploaded it to Grammarly for a quick check - only 4% AI patterns recognized.
You are right. There is no detection system available, and as soon as someone invents one, the AI bots will invent a tool to avoid detection.
You nailed it. I had a submission rejected last month from a popular online publication I hadn't worked with before--they said they don't accept AI written material. Don't know what detector they used, but it failed 100%
That is so frustrating. I'm sure that there will eventually be lawsuits against academic institutions (and now publications) that accuse students and writers of using AI.
Lots of good tells here, Maryan,
If you don't mind, I'll offer some perspective from a software engineer's standpoint, since I was doing that for years, too. Part of the reason current LLM models so frequently generate the kind of stuff you refer to is that current models are really just huge databases (really, really huge) of prior art.
AI processes are even distantly similar to some of the existing search tools that software people have been using for years, like SOLR, which derives its "intelligence" by matching patterns around letters so that when someone searches for music but types, "mustic" instead, it returns "music"-related stuff, because SOLR is aware of established character patterns in that word.
So AI is really just a big, fat old pattern-matching engine.
Their pattern matching is of such enormous complexity that there is controversy over the enormous amounts of energy they consume. It's too bad people insist on using it, because these things require massive data centers that directly conflict with goals for addressing climate change, but that's a different comment and post, I guess.
Yes, a different post--and I wish you would write it. Not many people know the toll this could take carbon-footprint-wise. Thanks for this insight--very interesting and completely worthwhile!
Thanks! Yeah, I have a few AI-related articles in my various queues here and on Medium. I haven't had a chance to flesh them out. I've run one or two, as well. Linda Carroll just posted an excellent one today.
Thank you, Maryan, for this interesting article. I'll be sure to pay attention.
This was so interesting, Maryan. Thankfully, I don't write like a Bot, I hope!
I'd like to know your opinion on using AI as an editor for spell check, grammar, and punctuation. I don't like Grammarly, never have. I have issues with using passive voice and that's mostly what I use it for. I don't know if using AI is good or bad for this, but I do use it for something other than writing, so pay for a subscription.
The one thing I notice when using AI when it comes to punctuation. It uses a lot of em dashes. I use a lot of them too, I like dashes and ellipsis, I write like I think. It's me and I'm not changing that, but noticed it likes to add them into places I wouldn't typically use them.
And for the record... my dad is rolling in his grave right now knowing I can't use the word 'plethora' - it was a word inside a family joke, used for years, and it makes me smile when I see it.
Loved this, Maryan. It was interesting to reading gives me things to think about.
Thank you, Patti - you're always so supportive. I write the way write, growing and polishing as time continues to go by. As for AI for grammar, I have used Gemini and GPT, but I prefer Grammarly to do it inline as I write--saves me time. I ignore almost everything Grammarly tells me except for actual rules of grammar stuff. I don't need a bot to refine my style. I set the settings in Grammarly to follow my rules. If you hate that one, you might try Quillbot--
I didn't know I could set the settings in Grammarly. I'll go back and revisit, maybe learning to use it might help. I'll check out Quillbot, thank you for the suggestion.
I use em dashes all the time. I've heard they are tells but I don't care. I like them.
I do too. And doing what I want, that's what having my own newsletter is all about... otherwise I'll write on Medium.
Elevate, unlock, and revolutionize are among many that are overused by AI and I've banned it from using in my custom instructions and memory (ChatGPT).
Thanks for sharing your list!
I find that AI writing sounds voiceless and generic. It reads like the instructions for opening a cereal box even as it’s allegedly about a personal experience. It reads without any variety in the sentence length. That’s a major AI tell over buzzwords that humans will use with impunity so long as someone who is winning at life is using them.
But equally, I don’t care if someone thinks what I write is AI because anyone who doesn’t trust themselves isn’t going to trust me and that says a hell of a lot more about them than me, so by all means, the tools are always going to assume that other people are tools and are using tools to cheat and squeak by just like they would. And those people give me ewe.
Haha, this is funny. I never thought about that. I like AI, but writing comes from my heart and it isn't going to take away my DNA thing for my doing so... if people don't want to read me, that's okay, too. The trust issue is interesting, glad you brought it up. Thank you.