Ai energy must be more efficient
Artificial intelligence uses too much energy. Developers need to find better ways and risks to add to the climate in the climate
Artificial intelligence is everywhere: it Design of new proteins, Answer Internet search questionsalthough Run barbecues. Investor fascinate that and is the US president. After taking office, President Donald Trump announced his support Stargate, a company worth $ 500 million, divided some of the largest players in this space to facilitate the development of AI In the US
But the data centers and other infrastructures needed to develop technology and other infrastructure are tremendous electricity pigs. And with Trump’s declaration of “National Energy Emergency”, it is able to increase the energy needs of fossil fuels production air to make the climate change even more serious. Technology is already in charge of Greenhouse gas emissions that cause climate change. Many other companies developed by Stargate and AI platforms insist on the energy cleaner and more efficiently, they will only help to destroy our planet.
Many of the technologies among many flavors, buzzy is knifeed, the basis of the chatGT and Google’s search response system. When performing operation, the AI creator must store the necessary computation in the two phases of electricity, the warehouse size data center.
To help Science Journalism
If you enjoy this article, consider entering award-winning journalism Subscribe. By purchasing subscription, you are helping to ensure the future of stories about the discoveries and ideas that are conformed to today.
We all will have to pay this exorbitant use of energy when it inflates the cost of Kilowatt, regardless of our personal interaction with technology.
The developers must first train the AI model in broad data stores, as long as there are many hours and requires horrible computer skills. A pioneering training dog was consumed enough electricity every year at the average US home. Each time a model is renewed, it must be reconsidered. A sudden liberation of depth prepared for the price division of our system and similar to our system can cause fewer energy processes, but it is sure to know it soon.
And the demand is not stopped when a model is prepared. Each consultation receives AI requires all feeds nied and then synthesized a response from scratch in the process called inferences, which also requires energy. Compared to search engines, the text generation system can easily use a query 10 times, and sometimes they use dozens more times. The generation of the image needs more energy, 50% of the amount needed to fully charge a smartphone, a study is found.
Many analyzes interpret this energy use of air owner’s training and large-scale operation as a higher cost of the system owner. For example, if Google uses a creator AI, in response only half of the questions he receives, the company will cost $ 6 billion.
It is true that we all have to pay for this exorbitant use of energy when it swells the cost of kilowatt, regardless of our personal interaction with technology. The scale of consumption is too large, and as ARO is included in more aspects of daily life, it saves its energy. On the scale of the industry, it is difficult to isolate AI from other computing requests, but data centers serve as a comfortable proxy, if the technology of technology has they moved to their boom.
The numbers are being staggered: in the mid-2010 decade in the US data centers a year in about 60 terawatt hours. (Terawatt-an hour is equivalent to millions of kilowatt, the unit was used Measure electricity consumption In most US homes.) 2023. For the year, the final report of the Lawrence Berkeley National Laboratory was found, almost triple 176 terawatt hours; The demand is expected to rise between 325 and 580 terawatt hours in 2028. At this level, the use of data center energy would be between 6 and 12% of US energy consumption, up to 4 percent in 2023.
The demand for commercial energy continues to grow, people are already watching Residence energy prices Thirst technologies like Alas in some regions that are taxed.
Among this Skyrocketing of energy demand, the work of downloading energy is slowly done in the US and the world. Climate change is already spreading around us, they are bending wrong disasters Los Angeles fires to Helene hurricane to Extreme heat and cause Amazing conclusions and long durations. Reducing damage to climate change requires the use of fossil fuels as soon as possible. Any tremendous demand for any industry is more difficult.
Sure, great technology companies Without the player Provide valuable resources to support energy transition. Stargate investment is expected to rely on a part of the solar power. Before leaving the office, President Joe Biden opened public land to data centers running in clean energy As a way to promote the use of computer science.
Because sun, wind and hydroelectric production rates can be changed with weather and other factors, nuclear energy is particularly attractive to prevent the AI technology companies, growing fears Nuclear waste pollution. Mainly Microsoft has an agreement to restart the three-kilometer island While the US is the worst nuclear accident, while Openai CEO Sam Altman is throwing his support, including nuclear fusion, Hardly energy should be given an important scale up to 2050 as soon as possible.
AI if companies smooth in clean power and do not bend the climate crisis, the apparent technology needs of technology. And efficiency improvements, necessarily, may not be enough. The so-called Jevon Paradoxa, being cheaper or more efficient can increase its use rather than reducing its trace. The wider highway invites more cars and the Internet has brought doomscrolling as a time consumer who encourages more energy use.
While technology companies promote AI, they do not only maintain small efficiency, but also the US energy footprint, but the alternative world can be character.
