Tech Beetle briefing IN

Why saying 'please' and 'thank you' doesn't really add to the cost of using AI

Essential brief

Key facts

Adding polite words like 'please' and 'thank you' to AI prompts has a negligible impact on computational cost and environmental footprint.

Significant environmental costs arise at the scale of billions of prompts, not from individual user interactions.

AI's environmental impact is more influenced by training processes, data center efficiency, and overall query volume.

Encouraging polite language can improve user experience without meaningful environmental trade-offs.

Efforts to reduce AI's carbon footprint should focus on large-scale operational efficiencies rather than prompt length.

Highlights

Adding polite words like 'please' and 'thank you' to AI prompts has a negligible impact on computational cost and environmental footprint.

Significant environmental costs arise at the scale of billions of prompts, not from individual user interactions.

AI's environmental impact is more influenced by training processes, data center efficiency, and overall query volume.

Encouraging polite language can improve user experience without meaningful environmental trade-offs.

In recent discussions about AI usage, some have suggested that including polite phrases like "please" and "thank you" in prompts to AI systems such as ChatGPT could contribute to increased environmental costs. This idea stems from the fact that AI models process text incrementally, meaning longer inputs require more computational effort. However, OpenAI’s CEO Sam Altman has clarified that while every additional word does contribute to operating costs, the impact is negligible when considered on an individual level. The scale at which these costs become significant is in the billions of prompts, far beyond typical user interactions.

AI models like ChatGPT operate by analyzing and generating text based on input prompts. Each token or word processed requires computational resources, which in turn consume energy. Naturally, longer prompts demand more processing power. Yet, the incremental cost of adding a few polite words is minuscule compared to the overall computational load. The environmental impact of AI is more heavily influenced by factors such as model training, data center efficiency, and the volume of queries processed globally.

The conversation about politeness in AI interactions also touches on human-AI relationship dynamics. While it may seem trivial from a cost perspective, encouraging polite language can foster respectful and positive user experiences. However, from a purely environmental and operational standpoint, the inclusion of "please" and "thank you" does not meaningfully increase the carbon footprint or energy consumption of AI services.

Moreover, the focus on reducing AI's environmental impact should prioritize larger scale efficiencies. Innovations in hardware, improvements in data center energy sources, and optimization of AI algorithms offer more substantial benefits. Users can contribute by being mindful of their overall usage patterns rather than the specific wording of their prompts.

In summary, while every word processed by AI systems adds to computational costs, the addition of polite phrases like "please" and "thank you" is not a significant factor in environmental impact. The broader challenge lies in managing and optimizing AI operations at scale to ensure sustainability. Users can engage with AI respectfully without concern for unnecessary environmental harm caused by their choice of words.