Latest news with #HochschuleMünchenUniversityofAppliedSciences
Indian Express
8 hours ago
- Science
- Indian Express
AI chatbots using reason emit more carbon than those responding concisely, study finds
A study found that carbon emissions from chat-based generative AI can be six times higher when responding to complex prompts, like abstract algebra or philosophy, compared to simpler prompts, such as high school history. 'The environmental impact of questioning trained (large-language models) is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions,' first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences, Germany, said. 'We found that reasoning-enabled models produced up to 50 times more (carbon dioxide) emissions than concise response models,' Dauner added. The study, published in the journal Frontiers in Communication, evaluated how 14 large-language models (which power chatbots), including DeepSeek and Cogito, process information before responding to 1,000 benchmark questions — 500 multiple-choice and 500 subjective. Each model responded to 100 questions on each of the five subjects chosen for the analysis — philosophy, high school world history, international law, abstract algebra, and high school mathematics. 'Zero-token reasoning traces appear when no intermediate text is needed (e.g. Cogito 70B reasoning on certain history items), whereas the maximum reasoning burden (6.716 tokens) is observed for the Deepseek R1 7B model on an abstract algebra prompt,' the authors wrote. Tokens are virtual objects created by conversational AI when processing a user's prompt in natural language. More tokens lead to increased carbon dioxide emissions. Chatbots equipped with an ability to reason, or 'reasoning models', produced 543.5 'thinking' tokens per question, whereas concise models — producing one-word answers — required just 37.7 tokens per question, the researchers found. Thinking tokens are additional ones that reasoning models generate before producing an answer, they explained. However, more thinking tokens do not necessarily guarantee correct responses, as the team said, elaborate detail is not always essential for correctness. Dauner said, 'None of the models that kept emissions below 500 grams of CO2 equivalent achieved higher than 80 per cent accuracy on answering the 1,000 questions correctly.' 'Currently, we see a clear accuracy-sustainability trade-off inherent in (large-language model) technologies,' the author added. The most accurate performance was seen in the reasoning model Cogito, with a nearly 85 per cent accuracy in responses, whilst producing three times more carbon dioxide emissions than similar-sized models generating concise answers. 'In conclusion, while larger and reasoning-enhanced models significantly outperform smaller counterparts in terms of accuracy, this improvement comes with steep increases in emissions and computational demand,' the authors wrote. 'Optimising reasoning efficiency and response brevity, particularly for challenging subjects like abstract algebra, is crucial for advancing more sustainable and environmentally conscious artificial intelligence technologies,' they wrote. PTI KRS KRS MPL
Mint
10 hours ago
- Science
- Mint
AI chatbots using reason emit more carbon than those responding concisely, study finds
New Delhi, Jun 19 (PTI) A study found that carbon emissions from chat-based generative AI can be six times higher when responding to complex prompts, like abstract algebra or philosophy, compared to simpler prompts, such as high school history. "The environmental impact of questioning trained (large-language models) is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences, Germany, said. "We found that reasoning-enabled models produced up to 50 times more (carbon dioxide) emissions than concise response models," Dauner added. The study, published in the journal Frontiers in Communication, evaluated how 14 large-language models (which power chatbots), including DeepSeek and Cogito, process information before responding to 1,000 benchmark questions -- 500 multiple-choice and 500 subjective. Each model responded to 100 questions on each of the five subjects chosen for the analysis -- philosophy, high school world history, international law, abstract algebra, and high school mathematics. "Zero-token reasoning traces appear when no intermediate text is needed (e.g. Cogito 70B reasoning on certain history items), whereas the maximum reasoning burden (6.716 tokens) is observed for the Deepseek R1 7B model on an abstract algebra prompt," the authors wrote. Tokens are virtual objects created by conversational AI when processing a user's prompt in natural language. More tokens lead to increased carbon dioxide emissions. Chatbots equipped with an ability to reason, or 'reasoning models', produced 543.5 'thinking' tokens per question, whereas concise models -- producing one-word answers -- required just 37.7 tokens per question, the researchers found. Thinking tokens are additional ones that reasoning models generate before producing an answer, they explained. However, more thinking tokens do not necessarily guarantee correct responses, as the team said, elaborate detail is not always essential for correctness. Dauner said, "None of the models that kept emissions below 500 grams of CO₂ equivalent achieved higher than 80 per cent accuracy on answering the 1,000 questions correctly." "Currently, we see a clear accuracy-sustainability trade-off inherent in (large-language model) technologies," the author added. The most accurate performance was seen in the reasoning model Cogito, with a nearly 85 per cent accuracy in responses, whilst producing three times more carbon dioxide emissions than similar-sized models generating concise answers. "In conclusion, while larger and reasoning-enhanced models significantly outperform smaller counterparts in terms of accuracy, this improvement comes with steep increases in emissions and computational demand," the authors wrote. "Optimising reasoning efficiency and response brevity, particularly for challenging subjects like abstract algebra, is crucial for advancing more sustainable and environmentally conscious artificial intelligence technologies," they wrote.
Time Magazine
14 hours ago
- Science
- Time Magazine
Some AI Prompts Can Cause 50 Times More CO2 Emissions Than Others
Whether it be writing an email or planning a vacation, about a quarter of Americans say they interact with artificial intelligence several times a day, while another 28% say their use is about once a day. But many people might be unaware of the environmental impact of their searches. A request made using ChatGPT, for example, consumes 10 times the electricity of a Google search, according to the International Energy Agency. In addition, data centers, which are essential for powering AI models, represented 4.4% of all the electricity consumed in the U.S. in 2023—and by 2028 they're expected to consume approximately 6.7 to 12% of the country's electricity. It's likely only going to increase from there: The number of data centers worldwide have risen from 500,000 in 2012 to over 8 million as of September 2024. A new study, published in Frontiers, aims to draw more attention to the issue. Researchers analyzed the number of 'tokens'—the smallest units of data that a language model uses to process and generate text—required to produce responses, and found that certain prompts can release up to 50 times more CO2 emissions than others. Different AI models use a different number of parameters; those with more parameters often perform better. The study examined 14 large language models (LLMs) ranging from seven to 72 billion parameters, asking them the same 1,000 benchmark questions across a range of subjects. Parameters are the internal variables that a model learns during training, and then uses to produce results. Reasoning-enabled models, which are able to perform more complex tasks, on average created 543.5 'thinking' tokens per question (these are additional units of data that reasoning LLMs generate before producing an answer). That's compared to more concise models which required just 37.7 tokens per question. The more tokens were used, the higher the emissions—regardless of whether or not the answer was correct. The subject matter of the topics impacted the amount of emissions produced. Questions on straightforward topics, like high school history, produced up to six times fewer emissions than subjects like abstract algebra or philosophy, which required lengthy reasoning processes. Currently, many models have an inherent 'accuracy-sustainability trade-off,' researchers say. The model which researchers deemed the most accurate, the reasoning-enabled Cogito model, produced three times more CO2 emissions than similar sized models that generated more concise answers. The inherent challenge then, in the current landscape of AI models, is to be able to optimize both energy efficiency and accuracy. 'None of the models that kept emissions below 500 grams of CO₂ equivalent achieved higher than 80% accuracy on answering the 1,000 questions correctly,' first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences, said in a press release. It's not just the types of questions asked or the degree of the answer's accuracy, but the models themselves that can lead to the difference in emissions. Researchers found that some language models produce more emissions than others. For DeepSeek R1 (70 billion parameters) to answer 600,000 questions would create CO2 emissions equal to a round-trip flight from London to New York, while Qwen 2.5 (72 billion parameters) can answer over three times as many questions—about 1.9 million—with similar accuracy rates and the same number of emissions. The researchers hope that users might be more mindful of the environmental impact of their AI use. 'If users know the exact CO₂ cost of their AI-generated outputs, such as casually turning themselves into an action figure," said Dauner, "they might be more selective and thoughtful about when and how they use these technologies.'
