Prime Highlights:

• OpenAI CEO Sam Altman notes that excessive politeness in AI interactions, including the use of “please” and “thank you,” is adding up to astronomical environmental and economic costs.
• Although these words increase user engagement, they introduce unnecessary computational overhead, resulting in increased energy usage.

Key Facts:

• OpenAI’s AI, such as ChatGPT, are based on data centers that use approximately 2% of all the power consumed globally, a figure guaranteed to increase as more AI is used.
• One 100-word email written by an AI takes 0.14 kilowatt-hours of energy, enough to light 14 LED lights for an hour.
• Altman proposes that there is a need to balance efficiency with the user experience, cognizant of the fact that AI has to be explainable as well as sustainable.

Key Background:

In a recent release, OpenAI Chief Executive Sam Altman expressed concern over the increasing energy consumption of AI, specifically that the overuse of courteous phrases in interactions with AI, i.e., “please” and “thank you,” might be driving up the cost of operations. These apparently insignificant words contribute to the computational work that needs to be done by AI systems in order to reply and thus consume more energy. Altman described how even though these sentences are a natural part of the manner in which humans communicate, they do have unforeseen effects when used with AI, causing increased energy usage.

The green footprint of AI is an ongoing concern, with data centers—which house AI services such as ChatGPT—consuming approximately 2% of the world’s electricity. That amount will grow as AI finds its way into additional industries. To put that into perspective, it takes the energy of 14 LED bulbs for an hour to generate a simple 100-word email using AI, a big number considering how many interactions with AI take place on a daily basis.

However, Altman realizes politeness in AI interaction is worth preserving if it will make the user experience better. Reducing verbosity may well lower environmental burdens, but in Altman’s opinion, such trade-offs should be made with care. The goal is to have AI systems remain productive and sustainable but not at the expense of relational nuances that make them more user-friendly and interactive.

Altman’s remarks speak to a wider issue confronting the technology sector: how to produce AI technology without causing damage to the environment. As demand for AI keeps on rising, it is essential that producers balance innovation in technology and protecting the environment.