Why Kindness Towards ChatGPT Costs Tens of Millions of Dollars?

Due to all its advanced features, launching AI is very expensive, requiring extensive capital for the underlying hardware and energy consumption. It turns out that engaging in casual and polite conversation with artificial intelligence leads to significant costs for companies like OpenAI. Nevertheless, the company’s CEO, Sam Altman, considers this investment worthwhile because, although these responses may seem trivial, using artificial intelligence makes it a bit more humane, reports Tom’s hardware.

Many of us use artificial intelligence daily, primarily for seeking assistance, but some users have established a deeper connection and converse with it as a friend. While artificial intelligence cannot express emotions, its human-like nature in these interactions instinctively drives us to use expressions of politeness such as ‘thank you’ and ‘please’.

Sam Altman confirms this and reports that ChatGPT spends tens of millions of dollars just on generating responses to these queries. Viewed differently, a recent report suggests that even a short three-word response from the LLM, ‘you’re welcome’, consumes approximately 40-50 milliliters of water. Keeping this in mind, it is evident that kindness contributes to OpenAI’s monthly costs, but this does not pose a problem for them.

Researchers from OpenAI and MIT suggest that some individuals may become emotionally dependent or even fully addicted to AI chatbots. This is likely to become more pronounced as AI conversations become indistinguishable from human ones, which will likely result in withdrawal symptoms, as is the case with any addiction.

Should We Be Kind to Artificial Intelligence?

It turns out that science supports ‘excessive’ digital kindness. Recent research has revealed a fascinating dual benefit of treating artificial intelligence systems with respect. Not only does it significantly improve the quality of AI outputs, but it also enhances our capabilities in ways we are just beginning to understand, reports Forbes.

Recent research has shown that politeness is not just about etiquette, but also measurably improves artificial intelligence performance. A study examining AI behavior in English, Chinese, and Japanese found that impolite instructions often resulted in poorer performance, including increased errors, stronger bias, and significant omission of important information.

When researchers tested different levels of politeness in tasks such as summarization and benchmarks for language understanding, they found that more polite instructions consistently outperformed impolite ones, with some models showing up to a 30 percent drop in performance for impolite queries. The research also revealed that different AI models respond to politeness in culturally nuanced ways, suggesting that, just like human interactions, cultural context is crucial in how we communicate with artificial intelligence.

Other research has uncovered an interesting phenomenon showing that individuals who maintain professional politeness in their interactions with artificial intelligence exhibit greater emotional intelligence in human relationships. It is not just about kindness but also about aligning social constructs in the brain for better leadership outcomes.

As artificial intelligence continues to tackle increasingly technical and analytical tasks, our unique human abilities—especially emotional intelligence and authentic connection—become more valuable. Therefore, it is essential that our interactions with artificial intelligence actually help us develop these critical skills rather than diminish them.