AI is now ruling the writing world — empowering content platforms while unfairly punishing real writers who have never even touched AI. This madness MUST stop NOW!
Thank you for writing this valuable and eye-opening story so clearly, Mike. Every word you said in this story deeply resonated with me as I experienced it firsthand and observed it clearly. We tested around 30 of these tools in 2022, and they all were wrong. So we stopped them. I don't understand why Medium still uses Copyleaks, which is the worst.
Your article hits the nail on the AI head. It's crazy when AI (a robot) asks you, a human, to confirm that you're not a robot. This world is becoming topsy-turvy.
One thing I'm leery of is I use tools like Grammarly to correct grammar and improve my text. I'm afraid the AI detection tools will start flagging all content as AI. I know a lot of people are frustrated with Medium. I hope they figure it out soon because I enjoy reading articles on Medium.
I would avoid its recommendations on "improving" text and rely on your skills as a writer. Grammarly is good at catching spelling errors and extremely obvious mistakes, but if you follow its recommendations for changing the structure of your sentences, your work will possibly be flagged, and not for the wrong reasons. If you're not confident in your sentence structure, send your story in anyway, and see what the editors say.
Mike – You never cease to amaze me! Thank you for putting into context why I am curious about AI and certainly have dabbled – but this piece also helps me better undersand and put into context – why I still prefer to start my research on my mind/body related matter with a good old-fashioned pen and paper in hand followed by a hard covered library book with the plastic still adhered.
I agree with your points, Dr Mike. These AI detection tools are so flawed.
A few weeks back, my sister who is a Phd holder, was preparing some proposal for submission (I think that's what she referred them) and when she ran her paper under AI detection (because now surprisingly even teachers check students submission under AI detection tools) and the results showed that it is AI generated. She was so stressed, saying that she had worked so hard in writing that paper and was clueless regarding how to proceed further because if she submitted that proposal, it will ultimately get rejected, even if she has written it by herself.
Also, sorry I wasn't able to read this story in Medium because I mostly read through my mobile and whenever I open Illumination publication to read any new story there, it only shows recent and old stories of 4 or 5 selected individuals. I am adding this point here because I suppose you are one of the editors there, so I should inform you about this issue. Even if I scroll for a long time still it is very difficult to find any new stories by writers apart from a few.
AI "learned" to write by scraping the internet etc, and emulating the good writing they found. Now they get credit for all that good writing. And good writers get penalized. And every piece of content submitted to the AI detectors is used to learn. What happens when coders try and teach AI to write? The "Humanizer." You nailed several great points in the article. Thank you.
And so have many others... what I don't get is how people are still being accused by these tools despite so much information about how they don't work.
I'm a Medium editor and writer. I can detect AI with my eyes closed. I don't need an algorithm. If someday AI writes with human introspection, I guess I'll have to just let AI write stories and essays, too. Right now, it can't.
One advantage I do have is that the pubs I edit vet writers before accepting submissions. If a pub does this well, they shouldn't need to worry too much about AI.
Illumination has a disadvantage because it has so many writers in its stable. So I don't know what its editors do when they're hit with the kind of volume of submissions that I assume they receive. Illumination acquired most of these writers before AI began hammering pub editors with submissions.
These days, if you're starting a pub (on Substack or Medium), you need to find a way to vet your writers, IMO, before accepting submissions from them. This can be as easy as asking them to write a few paragraphs about themselves. It's not that hard to detect AI-generated writing if you demand an angle.
I'd be curious to know what Dr. Yildiz says about the current approach by Illumination editors toward AI story floods. Do Illumination editors rely on their human instinct to manage this? Or do they use tools? I don't envy pubs with hundreds of authors.
Hi Charles, Dr Yildiz and his team tested these tools for 6 months with the support of 100 volunteer editors in 2022 and we unanimously agreed not to use them. We use human eye to detect them. It takes me and many editors only 30 seconds to spot AI generated content. We don't penalize writers for using Grammarly. Our readers have no issues with polished pieces as they don't like sloppy statements. Have you read his recent eye opening short blog on Medium? He stopped writing there for the last 3 weeks. https://medium.com/illumination/quick-update-on-recent-issues-6d979e5ef06a
That's great, Aiden. I don't think I'm quite that fast. The "with my eyes closed" may have been an exaggeration! :-) I thought I saw that story, but I didn't clap for it so maybe I didn't. Corrected. :-)
It is my opinion that we have finally arrived at the precipice of human endeavor. We set out, reaching for the stars, and then decided "F It!" and decided to just decorate the studio wall with pictures of stars.
After convincing ourselves that this reality was preferable to the real journey - we have set about to convince other people that our stars are the "real stars".
And it is not difficult. There are enough clinically stupid people around to outnumber and overwhelm the more intelligent of the species.
In fact, intelligence is a genetic aberration. Stupidity is the actual base-line "normal".
All that is left is for some "clever" pioneer to claim that we "can fly" as he/she plummets from the precipice - and for others to believe that falling IS flying.
(Which is true, in a sense. Albeit very short. And on a very tricky trajectory)
Our cunning endeavor and hubris has invited Nemesis to our door.
We can run - but Nemesis will eventually win. She always has.
Then we can go back to hand-drawing illuminated scripts in arcane books - barely aware of what they mean - but certain that the Priests know - even if we never will.
I have taken a break from Medium to work on my novel/poetry. All of this is very demotivating...ugh. I never saw income from medium though appreciated the interactions. Will need to rethink continuing on that platform.
It's a sad day for competent writers, who are being pushed off platforms to accommodate incompetence. Truly a dreadful situation, and unjust. Having to dumb down competence to bypass AI detection tools, that have no genuine regard for writing, is inhumane and unethical.
This was bound to happen. AI learns more quickly than the human brain can even imagine. As it learns to mimic human writing, and copies it, spreading its plagerism widely, what was originally penned by a person is now also generated by AI. Hence, the flagging of human expression as AI.
Thank you for writing this valuable and eye-opening story so clearly, Mike. Every word you said in this story deeply resonated with me as I experienced it firsthand and observed it clearly. We tested around 30 of these tools in 2022, and they all were wrong. So we stopped them. I don't understand why Medium still uses Copyleaks, which is the worst.
Your article hits the nail on the AI head. It's crazy when AI (a robot) asks you, a human, to confirm that you're not a robot. This world is becoming topsy-turvy.
One thing I'm leery of is I use tools like Grammarly to correct grammar and improve my text. I'm afraid the AI detection tools will start flagging all content as AI. I know a lot of people are frustrated with Medium. I hope they figure it out soon because I enjoy reading articles on Medium.
I would avoid its recommendations on "improving" text and rely on your skills as a writer. Grammarly is good at catching spelling errors and extremely obvious mistakes, but if you follow its recommendations for changing the structure of your sentences, your work will possibly be flagged, and not for the wrong reasons. If you're not confident in your sentence structure, send your story in anyway, and see what the editors say.
Mike – You never cease to amaze me! Thank you for putting into context why I am curious about AI and certainly have dabbled – but this piece also helps me better undersand and put into context – why I still prefer to start my research on my mind/body related matter with a good old-fashioned pen and paper in hand followed by a hard covered library book with the plastic still adhered.
TGIF brother! 🙏
I agree with your points, Dr Mike. These AI detection tools are so flawed.
A few weeks back, my sister who is a Phd holder, was preparing some proposal for submission (I think that's what she referred them) and when she ran her paper under AI detection (because now surprisingly even teachers check students submission under AI detection tools) and the results showed that it is AI generated. She was so stressed, saying that she had worked so hard in writing that paper and was clueless regarding how to proceed further because if she submitted that proposal, it will ultimately get rejected, even if she has written it by herself.
Also, sorry I wasn't able to read this story in Medium because I mostly read through my mobile and whenever I open Illumination publication to read any new story there, it only shows recent and old stories of 4 or 5 selected individuals. I am adding this point here because I suppose you are one of the editors there, so I should inform you about this issue. Even if I scroll for a long time still it is very difficult to find any new stories by writers apart from a few.
They just don't want to pay writers.
AI "learned" to write by scraping the internet etc, and emulating the good writing they found. Now they get credit for all that good writing. And good writers get penalized. And every piece of content submitted to the AI detectors is used to learn. What happens when coders try and teach AI to write? The "Humanizer." You nailed several great points in the article. Thank you.
I've also written about how these detectors don't work: https://medium.com/towards-data-science/accusatory-ai-how-misuse-of-technology-is-harming-students-56ec50105fe5
And so have many others... what I don't get is how people are still being accused by these tools despite so much information about how they don't work.
Thanks mate, I read it and left a comment. I am gonna feature that story this weekend. It deserves it.
Thanks, Mike!
Here’s an idea: let’s stop using AI detection and go back to just evaluating the writing as writing.
As more real humans take the time to use AI as a tool to assist and polish their writing - the more the detection software will be inconsistent.
Because a tool doesn’t do the job - the tool makes doing the job easier.
Just my perspective.
I'm a Medium editor and writer. I can detect AI with my eyes closed. I don't need an algorithm. If someday AI writes with human introspection, I guess I'll have to just let AI write stories and essays, too. Right now, it can't.
One advantage I do have is that the pubs I edit vet writers before accepting submissions. If a pub does this well, they shouldn't need to worry too much about AI.
Illumination has a disadvantage because it has so many writers in its stable. So I don't know what its editors do when they're hit with the kind of volume of submissions that I assume they receive. Illumination acquired most of these writers before AI began hammering pub editors with submissions.
These days, if you're starting a pub (on Substack or Medium), you need to find a way to vet your writers, IMO, before accepting submissions from them. This can be as easy as asking them to write a few paragraphs about themselves. It's not that hard to detect AI-generated writing if you demand an angle.
I'd be curious to know what Dr. Yildiz says about the current approach by Illumination editors toward AI story floods. Do Illumination editors rely on their human instinct to manage this? Or do they use tools? I don't envy pubs with hundreds of authors.
Hi Charles, Dr Yildiz and his team tested these tools for 6 months with the support of 100 volunteer editors in 2022 and we unanimously agreed not to use them. We use human eye to detect them. It takes me and many editors only 30 seconds to spot AI generated content. We don't penalize writers for using Grammarly. Our readers have no issues with polished pieces as they don't like sloppy statements. Have you read his recent eye opening short blog on Medium? He stopped writing there for the last 3 weeks. https://medium.com/illumination/quick-update-on-recent-issues-6d979e5ef06a
That's great, Aiden. I don't think I'm quite that fast. The "with my eyes closed" may have been an exaggeration! :-) I thought I saw that story, but I didn't clap for it so maybe I didn't. Corrected. :-)
Thanks for the reply.
My question is can you anyone make a clear distinction between AI Generated vs AI Assisted?
Good question. I think that depends on how much assistance you're talking about. If it's just some Grammarly changes, probably not.
I'm curious if the Medium editors are overwhelmed and are looking for ways to stay ahead of AI-generated content. I believe they are failing.
It is my opinion that we have finally arrived at the precipice of human endeavor. We set out, reaching for the stars, and then decided "F It!" and decided to just decorate the studio wall with pictures of stars.
After convincing ourselves that this reality was preferable to the real journey - we have set about to convince other people that our stars are the "real stars".
And it is not difficult. There are enough clinically stupid people around to outnumber and overwhelm the more intelligent of the species.
In fact, intelligence is a genetic aberration. Stupidity is the actual base-line "normal".
All that is left is for some "clever" pioneer to claim that we "can fly" as he/she plummets from the precipice - and for others to believe that falling IS flying.
(Which is true, in a sense. Albeit very short. And on a very tricky trajectory)
Our cunning endeavor and hubris has invited Nemesis to our door.
We can run - but Nemesis will eventually win. She always has.
Then we can go back to hand-drawing illuminated scripts in arcane books - barely aware of what they mean - but certain that the Priests know - even if we never will.
Anyway - I have to be popping along.
Have a nice day. (if that's possible. now)
You wrote this comment like a movie script. Every word deeply resonated with me.
Thank you for this, Mike. A lot to process here but very illuminating, to say the least.
I have taken a break from Medium to work on my novel/poetry. All of this is very demotivating...ugh. I never saw income from medium though appreciated the interactions. Will need to rethink continuing on that platform.
It's a sad day for competent writers, who are being pushed off platforms to accommodate incompetence. Truly a dreadful situation, and unjust. Having to dumb down competence to bypass AI detection tools, that have no genuine regard for writing, is inhumane and unethical.
This was bound to happen. AI learns more quickly than the human brain can even imagine. As it learns to mimic human writing, and copies it, spreading its plagerism widely, what was originally penned by a person is now also generated by AI. Hence, the flagging of human expression as AI.