The Hidden Costs of Politeness in AI Interactions
A recent question posed on X about ChatGPT, the AI-powered virtual assistant developed by OpenAI, sparked a fascinating discussion. The question was: “I wonder how much money OpenAI has lost in electricity costs from people saying ‘please’ and ‘thank you’ to their models.” While it may seem like a lighthearted inquiry, the answer reveals some serious environmental and financial implications.
Sam Altman, CEO of OpenAI, revealed that the cost could be in the tens of millions of dollars. This is due to the energy required for AI systems to process even the simplest of requests. The phrase “please” and “thank you” might seem harmless, but they contribute to the growing demand for computational power that fuels these advanced technologies.
Environmental Impact of AI
ChatGPT is not just an AI chatbot; it’s a powerful tool that consumes vast amounts of resources. Every time users ask it to write an email, analyze their Instagram profile, or plan their monthly budget, it uses significant energy. According to Business Energy UK, ChatGPT’s daily electricity consumption is estimated at 40 million kilowatt-hours, enough to charge 8 million phones.
But the energy consumption isn’t the only concern. Data centers, which are the backbone of AI operations, use water to keep their servers cool. It’s reported that these centers consume approximately 39.16 million gallons of water per day. That’s enough to fill 978,000 baths or flush a toilet 24 million times.
Why Does AI Require So Much Energy?
Generative AI, such as ChatGPT, is capable of creating content like text and images. This is achieved through large language models, which are neural networks trained on extensive internet data. However, this process requires massive amounts of power.
Morten Goodwin, a professor at the University of Agder in Norway, explained that data must be transmitted, processed, and stored, regardless of the complexity of the request. Even simple interactions, like saying “thank you,” require energy. This is also true for other online activities such as Google searches, emails, or Teams meetings.
Dr. Daniel Farrelly, a principal lecturer in psychology at the University of Worcester, highlighted that all online activity has a carbon footprint. Although individual impacts may seem small, when multiplied across billions of interactions globally, the environmental impact becomes significant.
The Role of Politeness in AI Interaction
Despite the environmental costs, many people still choose to be polite to AI. A study found that polite prompts have a negligible effect on how well AI performs. Neil Johnson, co-author of the study and a professor of physics at George Washington University, compared this to being nice to a toaster. He suggested that adding extra words like “please” and “thank you” can confuse the AI and increase costs for both users and companies.
Robert Blackwell, a senior research associate at the Alan Turing Institute, explained that AI processes words by tokenizing them into smaller pieces. The more tokens used, the higher the cost for companies running these models. Newer reasoning models often use even more tokens as they attempt to justify and verify their responses.
The Psychological Aspect of Politeness
While there are practical reasons to avoid excessive politeness, there are also psychological benefits. Research suggests that how we treat AI reflects our behavior towards others. Morten Goodwin, who is also deputy director at the Centre for Artificial Intelligence Research, noted that language models learn from user interactions. If users are polite everywhere, including with chatbots, it sets a standard for respectful communication.
Humanising AI: A Coping Mechanism
The tendency to humanise AI is partly driven by its ability to perform tasks at a near-human level. Luise Freese, a tech blogger, explained that this anthropomorphism helps people make sense of something that feels human but isn’t. She jokingly attributed this to pop culture influences like the movie Terminator, which has shaped public perceptions of AI as both exciting and potentially dangerous.
However, this humanisation can be misleading. AI systems do not possess thoughts or feelings; they merely mimic patterns. When people treat AI as friends, they risk forgetting that these tools are mechanical and lack genuine emotions.
The Risks of Over-Reliance on AI
There are growing concerns about the reliability of AI. Some chatbots generate false information, a phenomenon researchers have termed “hallucinating.” Medical experts have found instances where AI created fake health studies, while mental health professionals worry about people seeking therapy from these systems.
Ana Valdivia, a departmental research lecturer in AI at Oxford, pointed out that the tendency to humanise AI is not just innocent curiosity. It is often a result of how these technologies are marketed, encouraging emotional dependency or misplaced trust in systems that are fundamentally mechanical.
Conclusion
As AI continues to evolve, it’s essential to understand the hidden costs associated with its use. From energy consumption to environmental impact, the effects of our interactions with AI are far-reaching. While politeness may seem like a small gesture, it contributes to the larger conversation about responsible AI usage and sustainability. As users, we must balance our interactions with AI, recognising both its capabilities and its limitations.