Can ChatGPT’s writing be detected?
One of the questions that students and teachers frequently ask is: Can ChatGPT’s writing can be detected as being generated by a machine? This question is also often asked by writers, journalists, and content creators who are worried that their jobs may be taken over by AI in the future. The answer to this is not a straightforward one, but I’ll do my best to break it down in this article.
What is ChatGPT?
Before we dive into the main topic, let's first establish what ChatGPT is. ChatGPT is basically a super-advanced chatbot. It is a neural network that has been trained on a massive amount of data, including books, magazines, articles, and carious other types of text. It uses this data to write text that looks and sounds like I was written by a human. Sort of. Almost. It’s close but it’s not perfect.
Can ChatGPT's writing be detected?
The short answer is that it depends. There are certain situations where it may be easy to detect that the text was generated by an AI, while in other situations, it may be nearly impossible.
Some of the factors that can affect the detectability of AI-generated text include:
· Length: Short pieces of text, such as headlines or captions, may be more difficult to detect as being generated by an AI, as they are often formulaic and don't require much creativity. “This is a nice oven.” Does it matter if an AI wrote that? No. Nobody cares. But when it comes to long pieces of work, like an essay, then suddenly there’s some skin in the game and teachers want to know who the cheaters are.
· Complexity: Text that is highly technical or specialized may be easier to detect as being generated by an AI, as it may be difficult for the AI to understand the nuances of the topic.
· Tone: Text that requires a specific tone, such as marketing copy or persuasive writing, may be more difficult to generate convincingly with AI, as it may require a human touch. Especially if you’re using slang. ChatGPT’s writing is formal and it doesn’t know much about anything after mid-2021.
· Errors: AI-generated text may contain errors, such as grammatical mistakes or factual inaccuracies, which can be a dead giveaway that the text was not written by a human. ChatGPT will often invent things like imaginary source material, or fake people like doctors or scientists. If you are writing things with ChatGPT, then you need to fact check everything.
So, while it is possible to detect AI-generated text in some cases, in other cases, it may be nearly impossible.
Examples of AI-generated text
To give you a better idea of what AI-generated text looks like, here are some examples:
Example 1: Prompt: Write a short news article about a recent technology breakthrough.
AI-generated text: Scientists have made a ground-breaking discovery in the field of quantum computing. By using a new type of qubit, they have been able to perform calculations at a speed that was previously thought impossible. This breakthrough has the potential to revolutionize many fields, including cryptography and drug discovery.
Example 2: Prompt: Write a product description for a new smartphone.
AI-generated text: The new smartphone from XYZ company is the perfect device for anyone who wants the latest technology at their fingertips. With a stunning 6.4-inch display and an ultra-fast processor, this phone can handle anything you throw at it. Whether you're streaming your favorite shows or taking photos with the high-quality camera, this phone has you covered.
As you can see from these examples, AI-generated text can be very convincing and sound very similar to text written by a human.
How to detect if text was written by ChatGPT
While it may be difficult to detect AI-generated text in some cases, there are a few things that you can look out for:
· Repetition: AI-generated text may repeat certain phrases or patterns, as the model may have learned to generate these in certain situations. Most human writers will avoid using the same word over and over again. Editors will also force them to delete useless words or sentences like, “It’s important to remember that blah blah blah.” ChatGPT loves telling people who important things are. It also ends everything with, “In conclusion.”
· Unnatural phrasing: AI-generated text may contain phrases or sentence structures that don't sound quite right or don't flow naturally. Like warnings or suggestions that you should talk to a professional before using advice generated by ChatGPT. (They don’t want to get sued.)
· Lack of creativity: AI-generated text may lack the creativity or originality that is often present in human writing.
· Errors: As mentioned earlier, AI-generated text may contain errors or inaccuracies, which can be a dead giveaway.
· Lack of humor: ChatGPT isn’t very funny. It knows a few jokes, but you’ve probably heard them before, because that’s what ChatGPT does. It doesn’t create new information or have new ideas. It just rearranges and compiles text according the prompts that it is fed.
Can AI replace human writers?
While AI-generated text can be very convincing in some situations, it's important to remember that AI is not a replacement for human writers. There are certain aspects of writing that AI is not yet able to replicate, such as:
· Creativity: AI may be able to generate text based on a given prompt, but it cannot create truly original ideas or think outside of the box in the way that humans can.
· Emotion: Writing that is meant to evoke a certain emotional response, such as poetry or storytelling, is difficult for AI to replicate, as it lacks the emotional intelligence and understanding that humans possess.
· Context: AI-generated text may lack the context and understanding of the world that humans have, which can result in writing that is tone-deaf or insensitive.
While AI-generated text can be very convincing in some situations, it is not yet able to replace human writers. Human writers bring a unique perspective and creativity to their work that cannot be replicated by AI. However, AI can be a valuable tool for writers, helping to speed up the writing process and providing inspiration and ideas.
Why do teachers care if students use AI to help write their essays?
Teachers are starting to care if students used ChatGPT to help write their essays because it can potentially undermine the learning objectives of the assignment and the educational process as a whole. Here are some reasons why:
1. Academic dishonesty: Using AI to write essays can be considered a form of academic dishonesty if the student does not disclose it to the teacher. This is because the student is submitting work that they did not create entirely on their own, which goes against the principles of academic integrity. This is called plagiarism.
2. Lack of critical thinking skills: Essay assignments are often designed to encourage critical thinking and problem-solving skills. By using AI to write the essay, the student may not be engaging with the material in a meaningful way, which can limit their ability to develop these skills.
3. Limited understanding of the material: If a student relies entirely on AI to write their essay, they may not fully understand the material they are writing about. This can result in a superficial or incomplete understanding of the topic, which can lead to a lower grade or incomplete learning.
4. Missed learning opportunities: The process of writing an essay can be a valuable learning opportunity in itself. By working through the writing process, students can learn about research, analysis, and argumentation. If they use AI to write their essay, they may miss out on these learning opportunities.
5. Difficulty in assessing learning: If students use AI to write their essays, it can be difficult for teachers to assess the student's learning and understanding of the material. This can make it harder to provide feedback and tailor future assignments to meet the needs of the student.
So, while using AI to help write essays may seem like a time-saver or shortcut, it can ultimately be detrimental to the educational process. Teachers want to ensure that their students are learning and engaging with the material in a meaningful way, and using AI to write essays can sometimes prevent that from happening.
Some people are developing tools for teachers to detect if writing was created by ChatGPT, but I don’t think this is the right way to go about it. Calculators are accepted in classrooms because it makes the student more efficient. Some day, AI writing will accepted, and the learning process will change and we will test how competent a student is using a different method.