Shorter AI – Researchers behind a new UN-linked study say how people write to AI matters: dropping niceties like “please” and “thank you” could reduce ChatGPT’s energy use by up to 25%, saving tens of gigawatt-hours of electricity each year. The report ties everyday prompt
On a typical day, ChatGPT processes billions of questions—each one broken into tiny units of text known as tokens. Researchers at the UN University Institute for Water, Environment and Health argue the way people phrase those questions could change the energy bill.
Their report says cutting unnecessary words from AI prompts could reduce ChatGPT’s energy consumption by up to 25 per cent. The team calculates that removing “please,” “thank you,” and other niceties could save 87 to 98 gigawatt-hours of electricity per year. That amount, the report notes, is equivalent to the annual residential electricity use of up to 760,000 people in sub-Saharan Africa.
The message is careful, but not subtle. “We are not saying be rude to your AI. But don’t fall into the interaction trap and don’t go falling in love with it either,” said Kaveh Madani at UNU-INWEH.
Madani says concise prompts can save energy because they reduce both the number of tokens the model has to process and the number it generates in response. In some cases, he adds, shorter prompts may also simplify the task for the model—meaning less power is required overall.
The study is part of what it describes as one of the most comprehensive assessments of the environmental costs of AI to date. It warns that the environmental footprint of the technology is rising quickly, not just for electricity use, but also for land and water as adoption accelerates.
The scale of that adoption is staggering. ChatGPT alone processes around 2.5 billion queries every day, while Google processes 16 billion. Most of the latter include integrated AI summaries.
Yet even with that enormous demand, the researchers say companies disclose little about how much energy their systems consume. To estimate impact, they relied on available data from data centres.
The report says AI currently accounts for about 20 per cent of the energy used by data centres. That share, it projects, will double to around 40 per cent in the next few years. By 2030. AI alone could consume around 378 terawatt-hours a year. while data centres could use 945 TWh in total—almost 3 per cent of projected global electricity use.
Water is another pressure point. The researchers project that data centres could need 9.3 trillion litres of water by 2030—enough to meet the minimum annual domestic water needs of all 1.3 billion people in sub-Saharan Africa.
“We’re looking at something on a global scale that is being adopted faster than any other technology in the history of technology, so the energy use is increasing very rapidly,” Miriam Aczel at UNU-INWEH said.
In addition to urging people to change how they interact with AI, the researchers also call for changes from governments and industry. They say AI companies should be required to publish their energy consumption. Governments, they argue, should introduce energy caps on companies and individuals.
But the study’s most immediate target is everyday behaviour—because that behaviour feeds the machines. The researchers say people should be educated to use AI efficiently, avoid using it unnecessarily, and when they do use it, cut words and use less powerful models.
The energy gap between different tasks is also part of the message. The researchers say generating an image uses 60 times more energy than a text query. enough to power a 10-watt LED bulb for about 17 minutes. They add that a complex video uses up to 8000 times more than text. which could power the same bulb for about 1.7 days.
Madani’s closing comparison is blunt: “We are not saying AI is bad,” he said. “We are just saying let’s use it in a proper way. It’s like a knife: you can save a patient’s life in the operating theatre, but you can also kill someone with it.”
AI energy use ChatGPT prompts UNU-INWEH data centres carbon footprint tokens gigawatt-hours electricity water use by 2030
So… just don’t say please?? lol
I don’t know if this is even real, because people are gonna still use it like normal. Also energy use is like… not just the prompt, it’s everything else too. But I guess writing shorter can’t hurt.
Wait are they saying if you don’t use manners the AI uses less electricity? That feels kinda dumb like blaming the user for a tech company’s footprint. If ChatGPT is processing billions of queries, why isn’t the actual fix on their side like better servers or whatever.
The whole “don’t fall in love with it” line is weird. Like now we gotta be emotionally detached and also type less words? Also 760,000 people in Africa sounds made up-ish, like who measured that the exact way. I’m just saying if they really cared about energy they’d reduce the number of times it has to answer, not tell people to drop “thank you.”