Tools have become artificial intelligence Part of our daily life, whether in our help in writing articles and e -messages or explaining complex scientific theories. But what is often absent from us is the invisible price of these “technical miracles”, a price that the land pays from its energy, climate and resources.
Every time we ask a question to an artificial intelligence robot such as “Chat GBT”, a complex process begins behind the scenes inside huge data centers, sometimes the size of a football field, which depends on its operation on coal or gas stations, transforms our words into “symbolic definitions”, which are numbers used to address the demand, then pass into thousands of super -performing computers, according to a report of the “CN The American.
Using these super computers, the answer is born through dense mathematical processes that consumes an energy equivalent to 10 times what is required by a simple research on Google, according to an estimate of the American Electrical Electricity Research Institute.
How is the environment affected?
To find out the impact of this on the environment, researchers in Germany conducted a test that included 14 artificial intelligence systems that work with huge linguistic models “LLMS”, and asked them various questions.
The result is clear: the more complicated the answer, the more emissions Carbon dioxide. Detailed or analytical answers were responsible for emissions of up to 6 times what was generated by short answers.
As for the “smartest” models that have a higher conclusion capacity, such as those designed to answer complex legal questions, they have generated emissions more than 50 times more than their simplest counterparts.
Maximilian Doner, a PhD student at Munich University of Applied Sciences and the main author of the study, published in the magazine “The Prospects of Communications”, explains this disparity, saying, “The more accurate the model and its ability to think, the greater its energy consumption.”
He adds that these advanced models contain tens of billions of “teachers”, which are the biases that help the system to interpret and treat symbols, which makes them more able to analyze, but also more energy -consumed.
Appreciation is difficult
But what makes estimating the actual environmental impact of artificial intelligence is difficult is the difference in the different factors, such as the type of device used, the geographical location and the power source in the region. Therefore, the researchers preferred to use carbon emissions instead of accurate numbers, according to CNN.
The American network adds that the absence of transparency increases the difficulty of estimating the real environmental impact, explaining that most major artificial intelligence companies do not share accurate data about energy consumption, the size of their servers, or improvement techniques that they use, which makes it difficult for researchers and consumers alike to understand the true environmental cost.
Choli Rin, a professor of electrical and computer engineering at the University of California, confirms that every artificial intelligence model consumes resources in a different way, and must be evaluated according to the type of task in which it is used.
But at a time when companies rush to enter artificial intelligence tools in every application, location and platform, users may not have the final word for the extent or timing of the use of these technologies, according to CNN.
How do we relieve the effect?
Although user options may become limited as companies seek to link artificial intelligence with all applications, researchers suggest options to reduce the environmental impact of using artificial intelligence models.
Rin says that users should “choose wisely”, and Doner agrees with it, who says that the user must search for the model that fits his needs with accurately without consuming huge resources for no reason.
“Suppose you are a software engineer who deals with complex problems daily, then the use of an advanced artificial intelligence model may be a logical option. If you are a high school student looking for help to solve a simple home duty, resorting to these strong models is similar to using a nuclear calculator to perform a simple collection process.”
In his interview with CNN, he adds: If people were more aware of the average environmental cost to generate in response, they may have started thinking: Is it necessary to turn myself into a moving doll just because I am bored?