[소셜코리아 칼럼] Will the artificial intelligence explosion be ‘weak’ or ‘poisonous’ to the climate? Optimism-pessimism is tense… Must operate within the limits of the Earth’s ecosystem (Byung-kwon Kim, Research Fellow, Green Transition Research Institute) (⌚6 minutes)
The generative artificial intelligence craze brought about by Open AI over the past two years has finally begun to make waves in the area of climate response. There are many predictions that the artificial intelligence boom, which has raised all kinds of socio-economic optimism and expectations, can make a significant contribution to reducing greenhouse gases and preparing for climate disasters.
Greenhouse gas reduction vs. hippopotamus eating electricity
Received the Nobel Prize in Chemistry for developing the artificial intelligence protein structure prediction model ‘AlphaFold’ Demis Hassabis(Demis Hassabis) has already Last year’s artificial intelligence-based weather forecast model ‘Graph Cast’attracted attention by releasing it. The company he leads, DeepMind (known to Koreans as the company that developed AlphaGo, which played Go against Lee Sedol in 2016), has already developed an artificial neural network model in 2019 that predicts renewable energy generation much more accurately up to 36 hours in advance.
Graphcast weather analysis screen released by Google DeepMind. Google said, “Graphcast predicts weather up to 10 days in advance more accurately and faster than industry-standard weather simulation systems.” 2023.11.14. Provided by Google.
At the same time, the generative artificial intelligence boom is hyperscale data centerBy requiring enormous computer resources (accommodating more than 100,000 server racks), it has also created the stigma of ‘electricity-eating hippopotamus’. Hyperscale data centers typically have power demands exceeding 100 MW (megawatts). Annual power consumption is approximately the power required for 350,000 to 400,000 electric vehicles.It is equivalent to
As a result, as artificial intelligence improves and its scope of use increases, concerns are rapidly spreading that the impact on the Earth’s ecosystem will increase and the climate crisis may deepen to support it.
How should we accept this two-faced situation in which generative artificial intelligence prevents climate disasters and helps with efficient use of energy, but then consumes enormous amounts of computer resources and energy, adversely affecting the Earth’s ecosystem and climate? Should artificial intelligence get better and bigger for the climate, or should it be moderated and limited? This is a huge dilemma that our society is newly facing.
Optimism “Greenhouse gases can be reduced by an additional 10%”
Depending on experts or policy designers, approaches to this dilemma can be divided into optimistic, neutral, and pessimistic perspectives. First of all, we believe that generative artificial intelligence is an emerging new industry and an investment that will bring huge profits. Big tech engineers, executives and economists like Hassabis emphasize the positive net effect that artificial intelligence will have on climate response.Do it.
According to a report released by Boston Consulting Group and Google, “Artificial intelligence can be achieved by expanding currently proven applications and technologies. 5-10% of global greenhouse gas emissions can be reduced by 2030”, of course, argues that it has the potential to provide insights for preparing for and improving resilience to climate disasters. In particular, in the field of efficient management of distributed renewable energy, artificial intelligence is already used for more than 50 different purposes.
For example, a variety of data from smart meters and Internet of Things devices are used to observe and control the flow of power in the network, especially at the distribution level. In addition, artificial intelligence is used to optimally respond to electricity demand through electricity price prediction, response load scheduling and control, and dynamic price setting. Machine learning is also being applied in various ways to reduce greenhouse gases in the transportation and transportation field.
It raises issues not only about big tech engineers but also about the disappearance of jobs and violations of human rights. Even many experts who warn about the risks of artificial intelligence are relatively unconcerned about the ecological risks of artificial intelligence.all. This can be confirmed in the European Union’s ‘Artificial Intelligence Act’, which has led the way in regulating artificial intelligence, but even Europe has no provisions regulating the ecological risks that artificial intelligence can pose.
However, recently, as investments in artificial intelligence expansion and data center expansion have intensified, Big tech companies such as Microsoft and Google are facing disruptions in their carbon neutrality plans. When I throw it away, I sense a different atmosphere inside. This is the case when former Google CEO Eric Schmidt somewhat fatalistically claimed, “We’re not going to meet our climate goals anyway,” and rather than focus on reducing emissions, he wanted to “bet on artificial intelligence solving the problem.”
Neutral perspective, acknowledging both sides of ecological impact
Unlike engineers or managers in the big tech field, In the field of environmental policy or academia Take a step back from technological optimism and take a somewhat neutral position. Recognizing the ambivalence of artificial intelligence’s impact on climate and ecologyDo it. They remain optimistic in that they expect that artificial intelligence will help respond to the climate overall if it preserves its positive effect on the climate and reduces risks.
For example, Kim Sang-hyeop, co-chairman of the Carbon Neutral Green Growth Committee, pointed out that “AI is becoming a new main culprit in increasing greenhouse gas emissions by consuming enormous amounts of electricity,” but added, “On the other hand, AI is emerging as a leading player in climate energy solutions.” Acknowledge ambivalence.
However, as a solution, Chairman Kim proposed ‘Green of AI’, that is, making the energy required for AI as efficient and clean as possible, while ‘Green by AI’, that is, promoting greening through AI in earnest. However, it is not properly examined whether this process can control the total increase in energy and material use due to the explosion of artificial intelligence.
Similarly, a 2020 report from The European Policy Center found that Europe “Strengthening environmental protection and climate action through digitalization and simultaneously strengthening the greenness of the digital sector“I suggest we do it. The approach of maximizing the positive effects of artificial intelligence while regulating its negative impacts on the climate is quite attractive in itself. However, the challenges faced in concrete reality are not simple at all.
Criticism: Can the ‘Jevons Paradox’ be avoided?
Meanwhile, quite a few Environmental activists and ecological economists say the explosion of artificial intelligence poses a growing risk to climate and ecology.This will be seriously warningDo it. The danger primarily comes from excessive use of energy. It is already well known that generative artificial intelligence will rapidly increase the demand for energy and computer resources. For example, according to calculations by Professor Alex de Vries of the Free University of Amsterdam, if the current Google search function were implemented entirely with artificial intelligence, electricity demand would increase by more than ten times.
Of course, in developed countries and major economies such as China, the proportion of data centers today is limited to about 2-4% of total electricity consumption.And because the energy efficiency of AI semiconductor chips doubles every two and a half to three years, there are predictions that the impact will not be surprisingly large.
However, this is simply the global arithmetic mean. Specifically, as energy demand increases beyond expectations depending on the region and country, A movement to extend the life of existing coal and gas power plants or even to build new nuclear power plants and SMRs (small modular reactors).This is because it is appearing all over the world. In fact, data centers are concentrated In five U.S. states, more than 10% of total power supply is invested in data centers.In Ireland, an open country in Europe, data centers currently consume more than 20% of total electricity consumption, causing a serious burden.
So, what will be the net effect if we add the increased energy burden due to the explosion of artificial intelligence and the positive impact of artificial intelligence on the climate crisis, as mentioned earlier?
German ecological economists Steffen Lange and Johanna Pohl argue that, on the one hand, energy demand increases directly in the process of introducing, using, and disposing of digital technologies, and as a result, the overall economy grows, indirectly increasing energy demand. Demand was also expected to increase.
However, on the other hand, it was diagnosed that the introduction of artificial intelligence or new digital technologies could reduce consumption by optimizing energy use, and that the newly growing digital industry could replace existing carbon-intensive industries and reduce overall energy demand.
However, when the two conflicting trends come together, “Two increasing effects dominate the other two decreasing effects, so that overall, digitalization increases energy consumption.”The final conclusion was that they were doing it.
Although the study was conducted in 2020, before the advent of generative artificial intelligence, it is found that efficiency improvements actually lower prices and expand demand, so that the effect of increasing total usage outweighs the effect of reducing resource or energy use. ‘Jevons Paradox’can be applied to the digital and artificial intelligence fields without exception. Those who warn about the ecological harm of artificial intelligence emphasize that this is what we need to pay attention to.
Data center demand and impact must be accurately identified
In a situation where the competition for the survival of global big tech is expanding into a fierce battle to maintain digital competitive advantage between countries, it is not easy to stop the explosion of artificial intelligence development and application right now under any justification.
For this reason, artificial intelligence supporters come up with hundreds of reasons why artificial intelligence can help respond to climate change, provide evidence that the digital industry is a much lower carbon industry than other industries, or even, like Eric Schmidt, make some plans to reduce greenhouse gases. Even if they postpone it, they seem to have no intention of stopping, even listing the reasons why artificial intelligence competitiveness is so important.
However, at this point, it is necessary to heed the warning of Eamon Ryan, the Minister of Environment and Climate of Ireland, which is suffering the most from data center expansion. He said in an interview with the Financial Times on September 17th: “No industry can escape the climate challenges we all face.”He clearly points out that artificial intelligence and data centers must also “operate within the climate limits we promised.”
That’s right. Even artificial intelligence, whose life or death depends on the profitability prospects of powerful companies or the industrial competitiveness of a specific country, cannot escape the fact that it must operate within the limits of the climate.
Especially in Korean society, where dependence on the digital industry is high and optimism about artificial intelligence technology is widespread, the problem should not be avoided with the loose logic of simply ‘making artificial intelligence green and promoting greenness through artificial intelligence.’ For now It is urgent to accurately identify the demand for data centers and transparently and accurately evaluate the impact of energy increase and greenhouse gas reduction.do.