AI detectors are punishing human writers
So you spent some time, and put in thought, research, and effort to craft what you believe would benefit others, only to be accused of being AI content and pushing your writing down. If you have written an article or a story before you may have come across this accusation.
Writing has been an art for some, a form of expression for others. Some of us are full-time writers, some of us write as a hobby, and some of us write to put out an idea to the world. Some of us have a fully funded editorial team behind every writing and some of us love doing our solo writing.
Mission statement of medium
To deepen our collective understanding of the world through the power of writing.
Since the release of ChatGPT a lot of things have changed, we have seen a flood of AI-written articles without much thought put into it. The articles written for the sole purpose of benefitting only the author, without any clear start, inbetween, or end to the article. Soulless articles have caused huge problems not just for full-time writers but also for people who write to share their ideas or thoughts with the world.
Ideas vs Stories
Ideas stick words don’t
Ideas and words are two distinct thoughts, ideas stick but words don’t. So let's say, you read an amazing novel and when you are asked to tell the story to your daughter, would you recite the entire novel in the words of the author or do you tell the idea of the story in your own words?
Well you will tell it in your own words because you didn’t read the words, sentences, or paragraphs written by the author, you read the idea, that the author wanted to spread. The authors will make use of different captivating words to put that idea into you.
Let's take another example, but this time you read a scientific paper, so do you remember the definitions word for word or do you remember the idea?
So now if you pay for articles/novels/magazines do you pay for the idea or the writing? Well, that is not for me to say.
human editors vs AI editors
A lot of publishers including newspapers, magazines, and novels have fully funded editorial teams with multiple writers shaping the article's information, purpose, audience, and tone.
Almost all physically published books you have read have been reviewed and rewritten by multiple writers for the things stated above.
After the introduction of AI tools a lot of things have changed, some for good and some for worse. A lot of people now have access to the same or higher level of editing capabilities that weren’t easily available to common writers back in the day. It has helped a lot of people put their own ideas with a higher level of clarity and express their thoughts with a tone that hooks their audience, which tools like Grammarly have not been able to provide in the past without changing the context of the paragraph.
Without a doubt, it has helped me and many of you with the tone, spelling, and grammar of the articles we write to hook in a particular audience.
Non-Native writers
The introduction of Generative AI tools has come to help a lot of non-native English writers. This has changed the entire world of possibilities for them, they now have similar capabilities to spread powerful ideas in English, one of top most commonly spoken languages in the entire world.
Generative AI tools have not just helped non-native English speakers but also non-native French, non-native German, and others who want to express ideas in their non-native tongue.
Restricting the human writers
While the introduction of tools like ChatGPT has helped a lot of us. It has also limited our ability to use certain words that were commonly used in the past, not by AI but by humans ourselves.
A lot of common writing styles used by us in the past have been copied by AI, which in turn has started punishing human writers.
Many of us when young have used common phrases like “once upon a time…” to begin our story or “inconclusion” to conclude our writing. Phrases like this get fed into AI models which in turn start using these phrases in almost all of their writing styles, causing us human writers to not only mark it as AI writing but also causing us to stop using certain words for communication.
We are very close to the future where AI writing and human writing are going to be indistinguishable, in fact, we are living in it already.
AI today is human
Today AI has not reached a level of general intelligence (AGI), at least to that of the public knowledge as of January 1, 2024. So as long as we don’t reach there, AI will not be able to create its own ideas or work. Almost all work that has been produced by AI is just a collection of human intelligence provided in a presentable format.
Am I defending people blatantly using AI to generate content? Absolutely Not. Such people can go to hell! However, I will defend the use of these tools to help writers change the tone of the content and make their ideas presentable to their audience.
Problems with AI detectors
Now after a lot of blatant copy pasta from AI content generators, a lot of AI content detectors have emerged in the near past such as CopyLeaks, ZeroGPT, etc. Some of these tools have claimed to be about 99% accurate without warning users about inaccurate results.
Our AI engine has been trained to learn the patterns of human writing. Therefore, when the known patterns of human writing are disrupted, our AI flags it as potential AI content. No one wants to fear false positives that can lead to false accusations. We tested over 20k human-written papers, and the rate of false positives was 0.2%, the lowest false positive rate of any platform. In addition, we are continually testing our AI Model and retraining it with new data and feedback, helping to improve accuracy.
The above statement is taken from the CopyLeaks website, and the statement is AI content according to its own model.
Many of these AI detectors are also flawed and many ways. Let's take another example
Let's put this to the test, here are the results
Whoa! AI content right? Now let's make a very subtle change, we’ll just remove the space between subitles and subcontents.
And just like that it became human text. Many of these AI detectors also look for human errors such as flawed grammar, spelling, or even punctuation. But do you really publish an article with flawed grammar or spelling?
The worst feeling for any human writer is when their work is being inaccurately accused of AI writing, even when the author had put in research, thought, and effort.
The lack of warning users about inaccurate results by AI detectors have made a lot of people working in important positions at universities, companies, etc believe in the accuracy of these detectors.
So inconclusion, while some of these models may detect some of the AI contents correctly, companies, universities, and publications completely relying on these tools will only scrutinize human writers.
Embracing the AI future
Whether we like it or not, AI is here to stay. It is only going to get better from here. Using this as a tool doesn’t make you less of a writer, just like a doctor using a robotic arm doesn’t make them less of a doctor, or a mechanic using a tool doesn’t make them less of a mechanic.
A blatant copy-pasta without sharing knowledge is where we can say screw you!
Tackling AI content without scrutinizing human writers
Many universities, companies, and publications in the medium have started relying solely on AI detection tools without going into the article or understanding the content. The sole reliance on AI detection tools has caused a lot of problems for human writers.
We’ll discuss potential solutions for the Medium platform and how we can tackle content written purely by AI.
Marking AI articles
A simple prompt feature medium could be added before submission of the article so as to let authors know the medium platform of such use.
This could be used by medium for internal use and to refine their recommendations. Also, this could be a chance for authors to disclose if AI is used.
Stackoverflow/Reddit styles for medium
Humans are among the best detectors for AI-generated content, if content is generated by AI we’ll likely be the first to recognize it. An age-old moderation system by humans could be of great help. Reporting low-quality articles written by AI could improve recommendation feeds for many.
Not excluding writings
I believe that all writing has some value in it, some have more, so instead of taking out AI/AI-assisted articles from recommendation feed. Add weights to the articles, give higher weight to purely human-written articles, and push them even more. Ultimately let the human readers decide whether the article provides any value to them.
By combining one or more of the potential solutions we could greatly reduce false accusations, and presumptions and help human writers to write freely once again!
If you have other potential solutions or have been wrongfully accused of AI content, add a comment!
thanks for reading!
cheers
