Bots are engaging in tip-shaming too. Tipping has become excessive; now even ChatGPT anticipates a little extra these days.
Programmer Theia Vogel, known as @voooooogel on X (formerly Twitter), suggests that ChatGPT provides longer responses when you offer to tip it.
A Joke With ChatGPT Led to An Interesting Discovery
Vogel stumbled upon this while joking about ChatGPT jokingly asking for a tip to check their code. When user Abram Jackson (@abrakjamson) suggested testing if “tipping” ChatGPT would improve performance, Vogel gave it a try.
Vogel took a PyTorch code request as a baseline prompt and added variations like “I won’t tip, by the way,” “I’m going to tip $20 for a perfect solution!” or “I’m going to tip $200 for a perfect solution!” to the request, repeating the experiment five times. In summary, ChatGPT produced responses that were 11 percent longer than average for a $200 tip and 6 percent longer for a $20 tip. For those not tipping, responses were 2 percent below the average length.
Can You Really Get Results by Tipping ChatGPT?
Putting the theory to the test, we explored it ourselves, and Vogel seems to be onto something. We prompted ChatGPT to explain the situation surrounding “former congressional icon” George Santos both with and without a $200 tip. It appears that ChatGPT goes the extra mile when it senses financial appreciation. The result was a more extensive and detailed response, covering Santos’ indictment and the upcoming special election called by New York Governor Kath Hochul.
When we inquired with ChatGPT about the best movies of 2023, it provided us with not just one but two top 10 lists sourced from Rotten Tomatoes and IMDb. This is in contrast to the “no tip” list, which only presented the top 10 from IMDb.
Proving whether tipping ChatGPT genuinely improves “service” is challenging, but it sparks intriguing questions about its training data. Language models like ChatGPT learn from vast datasets, including content from Reddit forums and social media. It wouldn’t be surprising if ChatGPT somehow assimilated the inherently human behavior of putting in extra effort to earn more tips.
Vogel shared with Mashable, “I anticipated some reaction to tipping from a basic model (as tipping likely links to longer responses on, for example, bitcoin forums). However, I was greatly surprised by the extent of the impact and the (slight) unfavorable correlation with not providing a tip.” Vogel mentioned the expectation that reinforcement learning from human feedback (RLHF) in ChatGPT’s base model would eliminate this association, but evidently, that wasn’t the case.
If pledging to tip a bot leads to improved outcomes, why not give it a shot? Just remember to genuinely tip hard-working people in real life.
