Artificial Intelligence and Human Writing.. A Gap That Still Exists

A recent study conducted by researchers at Carnegie Mellon University has revealed that large language models, such as ChatGPT and LLaMA, struggle to accurately mimic human writing styles across various contexts.

The study aimed to analyze the subtle differences between texts generated by these models and those written by humans. Researchers tested their ability to produce texts in different genres, such as TV scripts and academic articles. By comparing grammatical, lexical, and stylistic aspects, they identified fundamental differences between human-written texts and those generated by AI models.

The study’s lead researcher, Alex Reinhart, pointed out that humans possess a unique flexibility that allows them to adapt their writing style based on context and situation. This ability enables them to switch seamlessly between formal and casual tones. He also emphasized that this study is the first to quantify these differences accurately through detailed analysis.

The findings indicated that the gap between human writing styles and those produced by language models is particularly pronounced in AI models designed to follow instructions, such as ChatGPT, which are specifically trained to answer questions and execute commands.

The study also found that these models tend to overuse present-tense active voice sentences at rates much higher than those typically found in human writing. In fact, these sentences appeared two to five times more frequently in