AI writing tools are freaking awesome. They can whip up texts on just about any topic, from fiction to non-fiction, with just a few words or sentences as input. They can help us with our writing tasks, like brainstorming, editing, and proofreading. They can also inspire us to be more creative and expressive with our words.
Sure, AI writing tools are pretty cool. But they can also be used for bad things, like cheating, deceiving, or manipulating people. They can produce texts that are not original, authentic, or truthful. This can make us question the credibility and validity of the texts we read online or offline. So, it’s important to be aware of the potential dangers of AI writing tools and to use them with caution.
How can we tell if a text is written by a human or an AI?
One possible solution is to use AI detection tools, which are supposed to identify and flag texts that are generated by AI models. These tools claim to use various techniques, such as looking at the language features, the logical coherence, the factual accuracy, and the stylistic patterns of the texts, to measure their probability of being AI-written. Some examples of these tools are GLTR, Real or Fake, and CopyLeaks.
Are AI Detection Tools Really Reliable?
According to a recent study by OpenAI, the answer is no. The study showed that these tools have a high false positive rate, meaning that they often mistake human-written texts as AI-written. The study also showed that these tools can be easily tricked by simple tricks, such as adding typos, changing words, or inserting irrelevant sentences into the texts.
The study concluded that AI detection tools are not trustworthy enough to detect AI writing and that they may even damage the trust and confidence of human writers. The study also warned that depending on these tools may create a false sense of security and that we should not underestimate the abilities of AI writing models.
This information alone should give you a good idea of why you shouldn’t trust AI detection tools completely. However, we will further explore the topic below.
What is Open AI’s Recommendation for Teachers?
OpenAI also said in the same source that AI detectors can’t always tell the difference between AI-generated and human-generated content. So, some teachers have started encouraging their students to share specific conversations from ChatGPT to show their work and formative assessment, demonstrate their ability to interact with AI and create accountability detectors that are not reliable in distinguishing between AI-generated and human-generated content.
This is because AI has become very sophisticated in recent years and can now generate text that is indistinguishable from human-written text. As a result, teachers have found it useful to encourage students to share specific conversations from ChatGPT to show their work and formative assessment, demonstrate their ability to interact with AI, and create accountability. This allows teachers to see how students are using AI and to provide feedback on their work. It also helps students to learn how to use AI responsibly and ethically.
AI has gotten really good at generating text that looks like it was written by a human. So, teachers are now encouraging students to share conversations they’ve had with ChatGPT to show their work, get feedback, and learn how to use AI responsibly and ethically. This helps teachers see how students are using AI and provide feedback, and it also helps students learn how to use AI in a way that is safe and respectful.
How can AI detection tools be misleading?
So, why shouldn’t you trust AI detection tools? here are some ways it can mislead you:
- They rely on superficial features: AI detection tools often use things like word frequency, sentence length, punctuation, and grammar to try to figure out if a text was written by a human or a computer. But these things aren’t enough to really capture the complexity and variety of human language. Plus, they’re easy for AI models or human writers to fake.
- They ignore the context and content: AI detection tools can be pretty dumb sometimes. They focus on the way the text looks, not what it means. They don’t think about the context of the text, like what the topic is, what the purpose is, who the audience is, or where it came from. So they might miss texts that are illogical, inconsistent, or inaccurate.
- They have limited data and knowledge: AI detection tools are only as good as the data they’re trained on. If they’re only trained on a limited amount of data, they might not be able to recognize AI-generated text that’s different from what they’ve seen before. This is why it’s important to keep training AI detection tools with new data so they can keep up with the latest developments in AI.
So, how can we make sure our work is high quality and accurate in the age of AI writing? One way is to use our own critical thinking and judgment skills, as well as other sources of verification, like plagiarism checkers, fact-checkers, and peer reviews. We can also use AI writing tools as a source of inspiration and help, not as a replacement for our own creativity and hard work.
How to Avoid Writing Like an AI?
- Use everyday language. This means avoiding jargon, acronyms, and overly formal language. Instead, use words and phrases that people would naturally use in conversation.
- Use first-person pronouns (I, me, my, we, us, our) and contractions (can’t, won’t, don’t, etc.). This will make your writing sound more natural and less like a textbook.
- Use active voice instead of passive voice. For example, instead of saying “The ball was thrown by the boy,” say “The boy threw the ball.”
- Use humor, anecdotes, and personal stories to make your writing more interesting and engaging.
- Proofread your work carefully to catch any errors in grammar or spelling.
As a takeaway from this article, we can conclude that though there are some tools available to detect AI writing, you shouldn’t trust AI detection tools fully as they follow certain patterns and can not fully assure that the detection is 100% accurate. It is best to use your own critical thinking and judgment to determine whether a piece of writing is AI-written or human-written.
If you want to learn more about what AI detection tools are available, and how they perform, you can check out our previous blog post on the best free AI content detection tools for teachers. We hope you find it useful and informative. Thank you for reading!
Khondker Mohammad Shah-Al-Mamun is an experienced writer, technology integration and automation specialist, and Microsoft Innovative Educator who leads the Blended Learning Center at Daffodil International University in Bangladesh. He was also a Google Certified Educator and a leader of Google Educators Group (GEG) Dhaka South.