Medium Is Testing a Tool to Detect If a Text Was Written by Artificial Intelligence
Listen up, AI writers. Bans and suspensions are what many readers and writers of Medium want.
It is clear the internet is increasingly being flooded with AI-generated text.
There are a lot of writers who have started writing their stories with *a little help*.
This little help means an AI = artificial intelligence.
Jasper AI for instance is one of their favorite tools.
Some clearly indicate that they are writing with the help of an AI.
Others don’t, although it’s often really clear that their texts are generated with the help of an AI… am I right?
This text was written by an AI — or was it?
Well, in fact, it’s not that easy to know whether a text is written by a human or a machine.
There are various ways to detect AI-generated text.
One common method is to use software to analyze different features of the text.
On Twitter, I saw, that Tony Stubblebine is testing a tool to detect if a text was written by AI.
“Green means the word had a high probability of occurring. One of these was written by AI and the other by a person.”
He added:
“I can’t tell the difference”
The software analyzes…
- how fluently the text reads
- how frequently certain words appear (this and that, it)
- whether there are patterns in punctuation or sentence length (AI tends to produce longer sentences with no mistakes).
Why does Tony want to know whether stories on Medium are written by AI?
“(…) I do know that I care to read it if it was written by a human and I don’t care to read it if it wasn’t.”
Plus, do you think it would be fair to continue to pay writers for text written with the help of AI?
A lot of writers and readers think AI-generated text should get banned and the accounts suspended.
To be honest, I think in reality it’s incredibly difficult to spot AI, and bans and suspensions are likely almost impossible to enforce (at the moment!).
Medium Top Writer Sean Kernan replied to Tony’s tweet and shared his opinion:
“I don’t worry about AI writing programs. They skew fluffy and tend to be impersonal, and nutty (IMO).”
However, isn’t it funny how Medium needs to use an AI to tell whether another AI wrote something?
This technological revolution presents us with a huge problem:
“How will we know whether what we read online is written by a human or a machine?”
“We’re in desperate need of ways to differentiate between human- and AI-written text in order to counter potential misuses of the technology”, stated Irene Solaiman, policy director at AI startup Hugging Face.
Users should clearly indicate text generated by AI
A spokesperson of OpenAI (known for their new ChatBot ChatGPT) shares that the company is working on watermarks.
Most writers who are looking for *a little help* don’t know about these policies.
For instance, OpenAI wants users to clearly indicate text generated by AI “in a way no one could reasonably miss or misunderstand.”
A problem I see for the near future
On Medium, there are a lot of non-native speakers from around the world creating texts.
Sometimes our texts (I’m a non-native from Germany) might not sound as fluid or conversational as others because we were taught another (more academic) writing style in school.
Additionally, we might use a lot of “this” and “that” (a typical sign of AI-written text) because we simply don’t know better and create a much cleaner text because we can’t write as we would in our mother tongue.
It could happen that when features of our texts get analyzed… the tool might think texts from non-natives are written by a machine.
The only way how to avoid it is to make stupid grammatical mistakes instead of checking your text over and over again!
They say about AI that if it has any virtues, it’s that they rarely misspell and make no grammatical mistakes.
Only joking!
