EUR
en
- es
- pt
- fr
- de
- it
Published on 2023-12-16
Artificial intelligence (AI) is a technology that demands a large amount of energy resources, both for its training and its use, and the more advanced it becomes, the more it needs, to the point where a query on ChatGPT consumes three times more energy than one made on Google’s search engine.
One of the most prominent companies in the current AI landscape is OpenAI, the creator of the 'chatbot' ChatGPT. It is estimated that up to 78,437 kWh of electricity was required to train the GPT-3 language model, a figure comparable to the energy consumption of an average household in Spain over 23 years.
This data, shared by the Institute of Engineering of Spain (IIE), is an example of the enormous energy consumption of artificial intelligence, as it usually requires large amounts of data for its training, while also needing a significant amount of computational and energy resources for its operation. And as AI models become more complex, their energy requirements can increase.
The consumption also translates into the liters of water used to cool the servers dedicated to powering the various products and systems of AI. Researchers from the universities of Riverside and Arlington estimated that training GPT-3 at one of Microsoft's most advanced data centers - a partner of OpenAI in the development of artificial intelligence - directly consumed 700,000 liters of clean fresh water, "enough to produce 370 BMW cars or 320 Tesla electric vehicles."
Google's own data centers in the United States, for their part, consumed 12.7 billion liters of fresh water in 2021 in the cooling processes of their computer equipment.
The use of this technology also has a high consumption level. For cooling, a single conversation of between 20 and 50 questions with ChatGPT requires about 500ml of water, according to the cited study. And in terms of electricity, a Google search consumes an average of 0.0003 kWh, while a query on ChatGPT consumes between 0.001 and 0.01 kWh, detail the experts from the Institute of Engineering of Spain.
"Large tech companies are driven by economic benefits and market demands, but they are also increasingly aware of the environmental impact of their operations and are taking steps to address this issue," explains José Andrés López de Fez, a member of the Digital Society Committee of the IIE, who cites the adoption of renewable energies to power data centers, the improvement of energy efficiency of servers, and the implementation of electronic product recycling programs.
More efficient chip design, research in the energy efficiency of AI that addresses not only hardware but also the efficiency in algorithms and model training practices, the use of renewable and carbon-neutral energy systems, improvements in data infrastructure, and electronic recycling and circular economy policies, are some of the issues that need to be addressed to reduce the consumption of energy resources.
COMMENTS
No customer comments for the moment.