- Blog
- A Test to Distinguish Between Human and AI Writing
A Test to Distinguish Between Human and AI Writing
Would you dare to take a test to distinguish which text was written by a famous essayist and which one by ChatGPT? This might just surprise you.
This time, we're going all in. Earlier, I posted a note with a poem generated by ChatGPT and shared it for feedback. The general opinion in the comments was that AI just piles on material, lacks substance, feels too scientific, and lacks emotional depth and creativity.
This time, I've decided to raise the stakes. Below are two paragraphs—one is from a renowned essayist, and the other is a piece generated by ChatGPT. For now, I won't reveal which is which. Take a guess, and see if you can figure out which one is from a human writer and which is AI-generated.
Why conduct this test? Because I've realized that many people, when judging AI-generated works, tend to carry a strong personal bias. No matter the content, if they know it comes from AI, it often gets dismissed immediately. In such cases, subjective emotions often replace objective judgment.
That's why blind testing becomes valuable. It reflects people's true reactions to AI-generated content. I must admit, generating a truly high-quality imitation piece with AI is no easy task. The literary capabilities of all large language models are still in the early stages. This is very different from solving a math problem or writing a simple program.
For example, asking ChatGPT to write a novel of 100,000 words is beyond its current capabilities. At present, large language models like ChatGPT can only handle shorter pieces, and even then, to match the ideal level of a professional writer, they require extensive revisions and adjustments.
Additionally, many people overlook a critical factor: to get AI to generate high-quality content, you must invest effort into crafting the right prompts. The quality of instructions greatly influences the quality of the output.
There's a common phenomenon where many users prematurely judge AI's abilities without thoroughly studying how to input instructions, which seems too hasty. It's like this: when faced with the same interviewee, an average person might not get much information, but a professional journalist can dig out plenty. High-quality questions and prompts are what unleash AI's true potential.
Alright, enough of the chatter. Let's get started. Which side will you pick? The answer will be revealed in a few days.
Excerpt 1
"The better part of Valour, is Discretion; In the which better part, I have saved my life. I do consider it is not the fashion to see the lady the epilogue, but a few words with a gentleman will do. However, I shall show more."
Excerpt 2
"To live in peace is to be truly wise; The world spins its ways, yet we must keep our gaze. For only in the quiet heart is there true power, And wisdom grows slowly, hour by hour."