Can you tell AI writing from human writing?

[Stuart Dren]

For some reason, stumbled upon this on Amazon as a book description. I scrolled down to see who translated the work, because it was NOT Longfellow as it claimed.

I think this is 100% AI writing. Because "Florence, which is present-day Italy" is so weird. Also, I have NOT heard of Le Rimes and Italian uses il and la! French uses le. I could be wrong. And it's weird that it lays out so many facts in a row. Also, Le Rime upon googling looks like either his work or a collaborative work. But still, haven't heard of it, or it didn't stick in my head. It also misses La Vita Nouva. His second most popular book.

What's your verdict? AI or not?

The low regard for paragraph structure made me think of Wikipedia, where it doesn't seem weird due to all the footnotes.

It's possible some copy-pasting, paraphrasing was performed by someone who doesn't write English well. The details themselves could have come from an LLM, but the broad structure or lack thereof doesn't seem LLM-like to me.

 

[Damage718]

Mods, delete if not allowed to post a new article or whatever reason.

I found this article interesting and I think it really shines a light on how humans make errors and why errors are important.

In this machine age we must hold on to imperfect writing. It is not flawed. It is human | Alex Reszelska

We need the mess of it all. Without it, what remains are sentences that are technically flawless but emotionally vacant

www.theguardian.com

Interesting article for sure. In a twisted way, it makes me feel (almost) better about a couple of my past books that have some minor typos and spacing hiccups littered in there. Clearly, human proofing (see: me) oversights. And truthfully, it makes me cringe that I missed them. To the point that I am actively cleaning up the manuscript to reupload. Anyhooo...

One book was written before the AI boom and the other was just at the start of it, so it's a flimsy excuse...but if goof-ups are now considered a sign of human writing, I'll take it

 

[Luxuria]

One book was written before the AI boom and the other was just at the start of it, so it's a flimsy excuse...but if goof-ups are now considered a sign of human writing, I'll take it

Agreed. I have dyslexia and now I feel a little less bad for missing words or having a mistake or two. Because then, my teachers really know it's not AI. I wrote all of it my self. I also keep confidence, I could write live and they would see it would produce similar results. Haha.

 

[Stuart Dren]

Agreed. I have dyslexia and now I feel a little less bad for missing words or having a mistake or two. Because then, my teachers really know it's not AI. I wrote all of it my self. I also keep confidence, I could write live and they would see it would produce similar results. Haha.

So there's a market for LLMs that mimick dyslexia then...

 

[Luxuria]

So there's a market for LLMs that mimick dyslexia then...

HAHAHHAHAAA. I can't wait.

 

[Pik]

The observation about the language being alive (in previous page) is interesting, I recently heard that the writing style in screenwriting is a bit flattened/dry. It's because the story only comes alive when filmed, it's not considered a literature or piece of art on its own. (Unlike e.g. plays that can be read like literature)

 

[FFBurwick]

I am sure AI is getting better and better at writing and creating art that are easily confused with what comes from human creativity. With regard to music, I even really like some of it that I found which in fact I know came from AI. I can't admire such that was produced from AI, which I can for human creativity. With writing now, it might take reading something at great length to become suspicious of it when it is from AI. I don't know what to think when something I posted someplace is judged as something a robot was posting. It has happened, though I think what was posted hadn't justified that. I have responded one time, "Yes, this is a robot, a robot just told you..." and I there list A, B, and C, of the points I had just communicated in what I previously posted.

 

[Pyroglyphian]

Mods, delete if not allowed to post a new article or whatever reason.

I found this article interesting and I think it really shines a light on how humans make errors and why errors are important.

In this machine age we must hold on to imperfect writing. It is not flawed. It is human | Alex Reszelska

We need the mess of it all. Without it, what remains are sentences that are technically flawless but emotionally vacant

www.theguardian.com

Me too. Thanks for sharing.

AI-writing is analogous to chronic cosmetic surgery, I think, whereby, no matter the starting point, people’s faces all seem to converge on the same hideous caricature.

 

[TrveKvltCatgirl]

By this point I can. You recognize the stock style - lots of specific verbs, wordage that sounds like ad copy, "yes man" cheeriness. AI writing can't ruin the em dash though.

Look, I'll admit that I dabbled in using gen AI to see what it would make of my prompts for the Hell of it. The dead giveaway was "rhyming" words that may have similarly spelled endings but in fact don't sound the same. "own" ---> "clown" for example.

I wrote better poems in middle school.

Is it weird that I almost miss when just a couple years ago when generative AI was new? When the most accurate AI-generated images were surrealist and psychedelic; when AI was just a novelty toy.

Doesn't make it any more ethical.

The scourge of AI on creativity has some unintended consequences past the ones everyone rightfully points out. Zao are an established metalcore band that have been paying dues their whole career. Their upcoming record is getting held up because an AI detection tool integrated with the digital platform used to distribute their tunes to streaming platforms detected a false positive in their unmastered songs.

As it pertains to writing, I think it especially doesn't bode well for novice writers or ones who grew up in sheltered, isolating circumstances, whose writing may be mistaken for gen AI.

Maybe one solution is artistic and deliberate subversion of grammatical rules and twisted formatting. Not exactly "House of Leaves" level. But one has to study the rules closely in order to break them with a purpose.

AI-writing is analogous to chronic cosmetic surgery, I think, whereby, no matter the starting point, people’s faces all seem to converge on the same hideous caricature.

I never looked at it this way, and this isn't the allegory I'd use, but you're spot-on. I'd liken the output more to indistinguishable, rotten coleslaw made by malfunctioning robot arms in a dark factory from the stolen recipes of tens of thousands of aspiring, talented chefs. Although I'd argue that some plastic surgery addicts are more sympathetic than those who rely on AI for anything from writing a bedtime story, making a tourism poster, to deciding where to go out for dinner.

Bottom line - we can't let human cognition itself be outsourced. The powers that be are invested in eroding the last vestiges of critical thinking skills in the masses. When someone said "I asked Grok / CGPT / etc." I just assume they're as gullible as a geriatric wowed by a Criss Angel stunt.

 

[Pending]

I'll admit that I dabbled in using gen AI to see what it would make of my prompts for the Hell of it.

Out of curiosity I've recently done and compared it against my writing. Is it bad that I think it might do it better? Not in every way, but the pacing seems to be improved. And how stuff is revealed...Dialogue also seems to deliver more information.

Ugh. I guess robotic competition has some benefits. Or maybe I should have just gotten human feedback. Woe is me.

Is it weird that I almost miss when just a couple years ago when generative AI was new?

I liked it too. It didn't feel like it was trying to deceive me, and it was fun with how bizarre it looked. Of course, it is sad that human surrealism was probably the first art form to be copied.

 

[Le gribouilleur]

Out of curiosity I've recently done and compared it against my writing. Is it bad that I think it might do it better? Not in every way, but the pacing seems to be improved. And how stuff is revealed...Dialogue also seems to deliver more information.

Ugh. I guess robotic competition has some benefits. Or maybe I should have just gotten human feedback. Woe is me.

I heard people being concerned when Deep Blue defeated Kasparov and when AlpaGo defeated the top go players. AI learns from people. We could learn from AI too.

 

[Stuart Dren]

I heard people being concerned when Deep Blue defeated Kasparov and when AlpaGo defeated the top go players. AI learns from people. We could learn from AI too.

For chess and other complex systems, sure. I don't want/need a middleman to learn from literature though.

 

[Homer Potvin]

Here's another one. Saw this over on .com....

https://www.thefp.com/p/ai-generated-literature-controversy?utm_campaign=realclearpolitics

Somebody may (or may not) have won a major word with something AI generated.

 

[Luxuria]

Here's another one. Saw this over on .com....

https://www.thefp.com/p/ai-generated-literature-controversy?utm_campaign=realclearpolitics

Somebody may (or may not) have won a major word with something AI generated.

Well, if this is not nightmare fuel, I don't know what is.

 

[Stuart Dren]

Between this and Hachette, the saddest thing is I'm not even surprised at how remiss these folks are. This one is sleeker than the suspected Shy Girl passages. It's subtly bad. It's the kind of bad that a judge would hopefully pick up on.

https://granta.com/the-serpent-in-the-grove/

At the clinic they cut the hem and wrapped the leg and checked for lights going out behind the eyes. The nurse had seen wells’ work.

Hehe. Not incoherent, but hilarious.

At the Mohammeds’ acre the light seemed thin. She saw the ring of stone, the lifted planks, the scuffed rope. She tore a length of vine from a mango trunk, peeled it in her hands to feel if it would hold a woman. She didn’t shout a name. She got to work.

That slip into pure LLM speak at the end is gold.

Wood complained in a voice too near speech. She lowered the pail until rope slackened. Smell rose – old wet, crushed jasmine, frog skin.

He grabbed the vine and hauled. Marsha hauled. Sita clawed stone. The well hated to give back what fell. Water is jealous. They pulled until Marsha’s shoulders were fire, until Vishnu’s hands were bone. Sita’s elbow hit stone, then her hips, then one knee. She slid, found a purchase that hadn’t been there a second ago and disappeared after. She came over the lip choking a sound the day almost refused. They lay on hot ground, breath scraping sky.

Night remade the house. The lamp smoked. Lizards hunted moths by the flame. Sita lay between sleep and pain, relief and a watchfulness that had nowhere to go. Marsha kept vigil. Vishnu stood by the broken mouth and didn’t go close. Bush took him in – not like a mother, like a judge. He had no words for the pressure on his chest, so the old names stepped forward: jumbie, duppy, serpent.

Sita said little but wasn’t quiet. She sorted beans with new slowness, looked at her boy with new exactness, and built inside herself a shelf for the decisions she would need to make when the time came. The shelf didn’t look like freedom – she couldn’t afford that word yet. It looked like not dying. It looked like not returning to a house where people forgot to see you.

A woman far off sang something too old for its words to matter.

Sprinkled nonsense.

It might seem like I'm nitpicking, but almost every paragraph is held together with shiny twine rather than craft.

 

[Rath Darkblade]

That's pointing out bad writing, not nitpicking. After I tried to read the first three examples, my eyes decided to skip the rest in self-defense.

Nicely done.

 

[Homer Potvin]

Nobel Prize winning author using AI for research apparently.

The zinger to me: "Contrary to fears, I believe that we writers, due to the specific nature of our craft, will most quickly and closely engage with tools like AI."

https://lithub.com/nobel-laureate-olga-tokarczuk-apparently-used-ai-to-write-her-latest-novel/

 

[defaux]

Nobel Prize winning author using AI for research apparently.

The zinger to me: "Contrary to fears, I believe that we writers, due to the specific nature of our craft, will most quickly and closely engage with tools like AI."

she believes readers are no longer interested in complex literary work

That's ... disappointing. Maybe she's given up?

 

[Rath Darkblade]

Nobel Prize winning author using AI for research apparently.

The zinger to me: "Contrary to fears, I believe that we writers, due to the specific nature of our craft, will most quickly and closely engage with tools like AI."

https://lithub.com/nobel-laureate-olga-tokarczuk-apparently-used-ai-to-write-her-latest-novel/

It depends entirely on which AI client you use and how you use it.

If you use the AI client and nothing else for your research, that's inadequate. If you then verify the results by reading further ... hmm.

If you use it to brainstorm ideas, that's OK. Not every writer has other writers to bounce ideas off of.

If you take the results of that brainstorming session and use them to write something in your own words that pleases you, that's fine.

But if you take what AI gives you and use it verbatim, that's obviously unethical. AI is trained on what is, to be blunt, theft. If you ask AI to write you a scene, and you take that scene and pass it off as your own work, you're an accomplice to theft. You're cheating yourself and your readers. There's no individuality, no spark of life, nothing that comes from you. You are misleading your readers, who then have no incentive to trust you. In short, you become a moral vacuum.

Thankfully, there are laws and policies in place in some countries to stop the practice of relying too heavily on AI. Australian law, for instance, specifically rejects legal exemptions that would allow AI companies to freely mine and scrape authors’ copyrighted work for training purposes without permission. But not all governments are that stringent; China, Russia, I'm looking at you. *taps foot sternly*

 

[Stuart Dren]

Thankfully, there are laws and policies in place in some countries to stop the practice of relying too heavily on AI. Australian law, for instance, specifically rejects legal exemptions that would allow AI companies to freely mine and scrape authors’ copyrighted work for training purposes without permission. But not all governments are that stringent; China, Russia, I'm looking at you. *taps foot sternly*

Sadly, if a company does scrape even without permission, do we generally have the means prove it? What's the worst they will face even if we can, a million dollar fine?