AI writing is developing extremely rapidly and showing better and better results. However, while illustrations created by AI can be easily labeled as produced by a specific AI model like Midjourney, determining the authorship of AI-generated text is more complex.
Currently, it is difficult to reliably recognize whether a text was written by AI or a human. Are there any reliable tools for this? Can ChatGPT 4 bypass them? Let’s explore these questions.
What can ChatGPT 4 do?
The capabilities of ChatGPT 4 (“Generative Pretrained Transformer 4”) are impressively extensive. In addition to working with text, it can now also process images. This fourth iteration of OpenAI’s software has undergone even deeper training, allowing it to generate text that sounds more human-like and provide detailed answers. However, like its predecessor ChatGPT 3.5, it is trained on data up to 2021. The creators, however, promise a significant reduction in incorrect answers, “hallucinations,” and biases:
“OpenAI claims to have spent the past six months improving the safety of the new software. According to them, ChatGPT 4 is more accurate, creative, and collaborative than ChatGPT-3.5, and it is “40% more likely” to produce factual responses.”
Is GPT 4 detectable?
The education sector is particularly concerned about the advent of ChatGPT, as it raises questions about students’ originality in essays. How can teachers determine whether a student has expressed their own thoughts or if ChatGPT has written everything for them? To save teachers and lecturers from spending excessive time checking for plagiarism and to discourage student cheating, educational institutions utilize reliable GPT 4 detector. One such detector is TraceGPT by PlagiarismCheck, which can distinguish between sentences written by humans and those that show clear signs of GPT, including the most recent release.