AI's progress linked to environmental harm
A new study from Germany found that big, advanced AI models—especially those that explain their answers step by step—can produce significantly more carbon emissions than simpler ones.
As AI gets smarter, its environmental impact is growing too.
Researchers tested 14 large language models (LLMs)
Researchers tested 14 large language models (LLMs) using 1,000 questions each.
Some high-performing models like DeepSeek (70B) used significantly more CO2 for every 500 questions, while others achieved similar results but with much less pollution.
Longer, detailed responses use up a lot more energy
The study found that longer, detailed responses use up a lot more energy.
Lead author Maximilian Dauner pointed out it's better to pick the right-sized model for each task instead of always defaulting to the biggest one.
AI's carbon footprint needs to be considered
With AI becoming part of daily life everywhere, these emissions add up fast.
The researchers hope people will start thinking about AI's carbon footprint and choose smarter ways to use these tools—so we can enjoy tech progress without hurting the planet.