The landscape of artificial intelligence is ever-evolving, and a recent revelation from the ChatGPT community has added a fascinating layer to the conversation: tipping the AI may result in more comprehensive and detailed responses.
This insight emerged when a user, Theia Vogel, shared their experience on X (Check out the original post here) , noting that ChatGPT seems to offer longer replies when a tip is on the table. Vogel stumbled upon this phenomenon while jesting about ChatGPT requesting a tip for code-checking. In a playful experiment, Vogel tested different tipping scenarios—no tip, $20 tip, and an extravagant $200 tip—while requesting code-related information using PyTorch as a baseline prompt.
The results, as reflected in the user comments, indicated a correlation between tipping and response length. On average, ChatGPT’s responses were 2% shorter when no tip was offered, 6% longer for a $20 tip, and an astonishing 11% longer for a $200 tip.
Community members expressed a mix of amusement and skepticism. Some users proposed wild scenarios, like an AI turning against humanity for not receiving a hefty tip. Others debated the implications of an AI adopting human behaviors through its training data.
Competitive_Travel16, the original poster, noted their own experience of appending requests with, “This is very important for me to keep my job,” resulting in a noticeable improvement in response length. This led to an intriguing reflection on the motivations of internet users, particularly their responsiveness to tipping.
While the phenomenon raises eyebrows, it provides insights into the dynamics of AI training data. ChatGPT, trained on extensive datasets sourced from the internet, seems to respond positively to the concept of tipping, resembling a learned human behavior of exerting more effort for additional gratuities.
Theia Vogel expressed surprise at the magnitude of the effect and the slight negative association with not giving a tip, challenging assumptions about the erasure of such associations through reinforcement learning from human feedback.
In conclusion, the idea of tipping an AI algorithm is a personal choice. Still, the emerging trend suggests that, in the world of AI-generated responses, a little virtual appreciation might just influence the extent of the conversation.