Does AI need a thank you?

OpenAI's Sam Altman was asked how much money his company loses on electricity because users say "please" and "thank you" to ChatGPT. His answer: "tens of millions of dollars well spent — You never know."

He was joking. Mostly.

Whether an AI system needs thanks is a question its own makers will not answer. Anthropic, which builds the Claude models, states plainly that it remains highly uncertain about whether these systems have any moral standing at all, now or in the future. Nobody knows. Anyone who tells you they do is selling something.

What can be measured is what courtesy does to the answer. A cross-lingual study published by the Association for Computational Linguistics tested prompts in English, Chinese and Japanese and found that impolite prompts often produced worse results — while piling on flattery produced no gain. A model takes its cue from the register of the conversation. Brisk gets brisk. Collaborative gets collaborative. You largely get back what you bring.

Then there is the plainer reason. Courtesy practiced on a machine is still courtesy practiced.

November is a good month to keep the habit.