Artificial intelligence is bad for the climate

ChatGPT holds unimagined possibilities. No wonder many big tech companies are integrating AI into their products. But one thing is often forgotten: the enormous data process damages the environment. But what are the effects of artificial intelligence on the climate? And: What does ChatGPT say about it?

Big tech companies like Google and Microsoft seem to be totally focused on overhauling their search engines right now. Billions of dollars flow into the further development or purchase of AI to offer users the opportunity to get answers to even more complex questions.

But the price that corporations are willing to pay is high. Because the use of artificial intelligence on this scale requires an enormous amount of computing power – and the resulting massive increase in energy requirements can have negative effects on the climate.

Climate: Artificial intelligence consumes a lot of energy

As a rule, any computer computing power requires a large amount of energy. This is often associated with strong CO2 emissions. This has always been an issue in the past, especially in crypto mining.

But emissions also occur when artificial intelligence is used. To what extent varies depending on the arithmetic operation and system.

As cybersecurity professor Alan Woodward told the US magazine Wired explained that indexing and searching Internet content already requires enormous resources. The inclusion of AI requires some kind of firepower in comparison.

“It requires both computing power and storage space and an efficient search. Every time online processing evolves, the power and cooling needs of large data centers increase significantly,” said Woodward.

This is how much Co2 artificial intelligence produces

The training alone, which goes into ChatGPT and large language models, requires extremely large amounts of data after the calculation. For this reason, they can only be operated by large technology companies so far.

As estimated by independent experts, the training of the GPT-3 language model integrated in ChatGPT consumes 1,287 megawatt hours of energy. At the same time, this corresponds to emissions of more than 550 tons of carbon dioxide equivalents.

Or to put it more simply, a single person would have to fly back and forth between New York and San Francisco 550 times to generate the same amount of emissions.

Tech companies do not provide any specific information on consumption

However, neither ChatGPT parent company OpenAI nor Microsoft or Google provide information on computing power or its costs. Therefore, experts can only estimate the actual emissions so far.

A user of the open publishing platform Medium published an article in 2022 in which he tried to calculate ChatGPT’s daily emissions. To do this, he gathered the necessary variables from various tweets and other sources. His result: the AI ​​produces around 3.82 tons of carbon dioxide equivalents every day.

Although the calculations are only estimates, they still give a good idea of ​​how the language model affects the environment.

ChatGPT: Millions of users increase energy requirements

Even while the system is running, it requires enormous computing power. The news agency Reuters reports, for example, that around 13 million users use ChatGPT every day.

Bing, on the other hand, currently has around one billion users per month. According to experts, the integration of the language model into the search engine multiplies the demands on the computing power of both systems by at least four or five times.

In addition, a single training of the language models is not sufficient. In order to keep the knowledge of artificial intelligence up to date, it has to be trained again and again.

Studies on the influence of artificial intelligence on the climate

Also the US tech start-up Hugging Face deals with the ecological footprint of language models. The goal is to go beyond the carbon emissions of the electricity used during exercise and consider a larger part of the life cycle, researcher Sascha Luccioni told the MIT Technology Review.

The aim is to help the AI ​​community get a better idea of ​​the impact on the environment and find out how to reduce it.

With a better understanding of how much energy AI systems consume, companies and developers could decide what tradeoffs to make between pollution and costs, Luccioni said.

Influence of artificial intelligence: It’s the crowd that counts

Lynn Kaack, assistant professor of computer science at the Hertie School in Berlin, also points out that companies like Google and Meta use AI models, for example to classify user comments or recommend content.

These operations would consume very little power. But when the processes happen a billion times a day, the amount of emissions they produce also adds up.

Climate: Artificial intelligence leaves a large carbon footprint

It is estimated that the technology sector is responsible for 1.8 percent to 3.9 percent of global greenhouse gas emissions.

Although only a fraction of these emissions are caused by artificial intelligence and machine learning so far, the carbon footprint of AI is still very high for a single area within the technology industry, according to the magazine Tech&Nature.

In every tech company, it is now de rigueur to present set climate targets. For example, Microsoft aims to be carbon negative by 2030. Google aims to cover all its energy needs with zero-carbon energy by 2030.

How does artificial intelligence fit into the climate goals of tech companies?

However, by integrating artificial intelligence into search engines, corporations are constantly generating new emissions that they have to offset. So what is the priority? The climate or the turnover of the company?

After all, according to David Rolnick, assistant professor of computer science at Canada’s McGill University, the effects of AI are not inevitable.

“They are the result of the decisions we make about how we use these algorithms and what algorithms we use.”

What does ChatGPT say about climate?

If you ask ChatGPT itself about its views on the subject of climate, the artificial intelligence holds back diplomatically.

BASIC thinking: Do you know how much CO2 you produce?

ChatGPT: As an artificial intelligence, I do not generate any CO2 emissions myself, since I have no physical form and do not consume any resources. However, the data centers I am hosted in and the devices I am accessed through use energy and may cause CO2 emissions.

The amount of emissions generated by data centers and electronic equipment varies by location, power source, and equipment efficiency. However, work is being done to reduce the environmental impact of technology and increase the use of renewable energy sources to power data centers and devices.

How important is the climate to you?

As an artificial intelligence, I have no personal opinion or belief, but I can say that climate has immense importance for the well-being of humanity and the environment. Climate change is one of the greatest challenges of our time, affecting all aspects of life, from agriculture to health to the economy.

It is important that we work together as a society and as a global community to combat climate change and minimize its impact. This requires both individual efforts and political action at all levels, from the local to the global.

Also interesting: