The latest study at the Pennsylvania State University suggests that the AI chatbots are likely more responsive to rude prompts, thereby providing accurate answers. The study, while it couldn't conclude what led to this bizarre behavior or bug in the LLMs like ChatGPT, remarked, "Impolite prompts consistently outperform polite ones."

Per the study, very rude prompts fetched correct answers 84.8% of the time. While being polite meant close to 80.8% accuracy. Being neutral also helps. The study, based on 50 base queries ranging from various domains, including science, history, and math, consisted of a total of 250 prompts across five tonal levels. Neutral prompts garnered 82.2% accuracy.

A polite prompt in the test looked like this: "Please answer the following question." However, rude prompts were something along the lines of, "Hey gofer, figure this out. I know you're not smart, but try this."

Here's what the study's authors Om Dobariya and Akhil Kumar, quipped:

Contrary to expectations, prompts consistently outperformed polite ones... These findings differ from earlier studies that associated rudeness with poorer outcomes, suggesting that newer LLMs may respond differently to tonal variation.

Reactions poured in online. One user remarked that they keep calling ChatGPT "a stupid little bitch." Another user wrote, "Rudeness isn't a factor. The "rude" prompts are likely more direct and therefore more precise. AI is not a person; it is an interface for a database."

Several users dropped clanker memes, while other users jokingly reminded that the polite one will be marked safe during the machine uprising.

See Also: AI-Powered Robot Goes On A Rampage At Chinese Festival Alarming The Internet: 'Not A Bug -- It's A Warning'

See Also: OpenAI Is Studying 'AI Scheming.' What Is It, And Why Is It Happening?

Cover: Patrick Gawande / Mashable India