An argument often used to quell growing concerns. Energy and resource demand Of data centers That is, artificial intelligence (AI) models will be needed less in the future as they get better and become more efficient.
But according to one, this seemingly logical thinking is a trap new united nations report Which quantifies the environmental costs of AI.
The report estimates that by 2030, AI energy use could double, consuming 3% of the world’s electricity, generating emissions equivalent to those of the UK and requiring more water for cooling than the annual drinking water requirement of the global population.
It is also predicted that the use of AI will follow an economic principle called the “Jevons paradox”, which predicts that when technological improvements increase the efficiency of a resource, it will lead to an increase rather than a decline in the total consumption of that resource.
The paradox is named after the economist William Stanley Jevons Who observed this effect from the use of coal in 19th century England. Efficiency gains did not reduce overall consumption. Instead, lower costs resulted in expanded use and higher overall demand.
As AI models become cheaper and more attractive, the report expects this will lead to new uses and greater amounts of use, which will reduce and possibly eliminate any savings from advances in efficiency.
To avoid falling into this trap, it lays out a roadmap for responsible AI use based on the guiding principles of transparency, efficiency by design, equity and justice, lifecycle responsibility, global collaboration and sustainable use.
Get the world’s hottest discoveries delivered straight to your inbox.
scale of the problem
Last year, data centers were already consuming as much electricity as Saudi Arabia Ranks as the 11th largest electricity consumer in the world.
If electricity use doubles by 2030 as projected, the associated carbon footprint would require 6.7 billion trees grown over ten years to meet this demand.
The data centers will also require 9.3 trillion liters of water and land approximately ten times the size of Mexico City.
Beyond resource use, the report also outlines the structural inequality at the heart of the AI boom, with only 32 countries hosting AI-specific cloud infrastructure and 90% of that capacity located in the US and China.
It warns of a growing digital divide between the countries that create and control AI systems and the countries that consume them, which often bear a disproportionate environmental burden caused by mineral extraction and e-waste.
Responsible AI Use
Two main forces shape the operational footprint of AI: how much we use it and how we use it.
This includes all the tasks performed by AI models, from text and code generation to images and videos. Each of these tasks requires different levels of computational effort.
The choice of model also matters because each AI system performs these tasks with different energy and environmental costs.
The report argues that responsible AI requires full value-chain governance, from mineral sourcing to recycling and safe disposal.
This calls for linking efficiency and environmental stewardship – thinking both about what AI can do for us and the protection of the natural environment.
This would mean making environmental disclosures a regular part of AI development at both the model and task level and incorporating anticipated AI demand into climate and energy planning.
Responsible AI is important as countries are promoting and adopting AI in government and the public sector.
In Aotearoa New Zealand, the government has launched a National AI Strategy and a Public Service AI Framework.
While the outline information was given OECD’s Values-Based AI PrinciplesIncluding inclusive and sustainable development, there are no requirements for environmental disclosures and no regulator to compile energy use or emissions.
Similarly, in Australia too, improving public services is part of it. National AI Plan. For example, the National Film and Sound Archive of Australia has created bowerbirdA machine learning-enabled mass audio and video transcription engine for document content. The Department of Veterans Affairs has Developed a proof-of-concept tool To see if AI can help speed up claims processing.
Both countries deliberately take a “light touch” and principles-based regulatory approach to AI. But this approach risks ignoring the growing environmental costs of AI that cannot be solved by improving it.
The natural environment is the basis of economy, culture and prosperity. This should be at the center of our thinking. Now is the time to rethink the AI innovation playbook and focus toward a sustainable technological future.
This edited article has been republished Conversation Under Creative Commons license. read the original article.