We need net-zero AI emissions — a global issue

• opinion Written by Philippe Benoit (Washington DC)
• Thursday, April 25, 2024
• interpress service

However, AI requires large amounts of computing power powered by electricity, which can generate additional emissions.

Unfortunately, climate modeling by the International Energy Agency and others shows there is no room for adding new sources of energy emissions. Therefore, to sustain climate change goals, AI must achieve net-zero emissions.

AI can reduce emissions in numerous activities across a variety of sectors. For example, AI can help reduce emissions from manufacturing, food systems, and road transport while increasing the production of zero-carbon electricity from solar and wind power.

However, recent reports point to a surge in demand for increased electricity production, particularly in the United States, driven in part by AI's computing needs. As a result, emissions are also increasing.

Additionally, with Nividia announcing the latest and most powerful AI performance chip, the growing demand for AI and the power required for AI may further increase. And this will happen not just in the US, but around the world, as reflected in Saudi Arabia's $40 billion investment plan in AI.

While there may be some efficiency gains thanks to AI, a net increase in electricity demand is expected, especially as the entertainment industry and others develop new and creative uses for AI.

This AI-driven increase could begin in the next few years, long before the power grid is transformed from its current fossil fuel-based system to a low-emission, renewable energy-based system.

As a result, emissions from the power sector are expected to increase further in the short term. In the long term, the available carbon budget (future emissions that stay within internationally agreed temperature targets) will also be reduced.

Climate scenarios similar to the IEA's 2050 climate scenario are built on the balance between carbon dioxide emissions and carbon removal from the energy sector. Deploying renewable energy, energy efficiency, fuel switching and other low-carbon technologies are key to reducing emissions to a level that is balanced by removal.

But unmanaged AI could hinder this effort, as its thirst for electricity could eliminate new sources of emissions. AI must “pay its price” on the climate by achieving net-zero emissions, and preferably even net-negative emissions.

How can we achieve this? First, a concerted effort to power data centers and other AI-related infrastructure with renewable energy without cannibalizing low-emission power generation projects for homes and other consumers. is required. Governments and the private sector (including local and remote data center users and private capital) must work together to increase investment in renewable energy and accelerate adoption.

A second possible tool would add heavy computational surcharges to AI users and possibly large-scale computational activities such as crypto mining to help finance additional investments in renewable energy. That's it.

Third, the climate impact of AI should also be added to the discussion about managing its possible negative impacts, such as misinformation and the potential for disruption to the job market. For example, the United Nations recently approved a US-led resolution to make AI “safe, secure, and trustworthy.” Similar efforts need to be scaled up to address the potential emissions impacts of AI.

Fourth, and perhaps most effective, we need to find mechanisms to activate the AI ​​itself to achieve net-zero emissions and enable net-negative emissions. This should include the development of innovative emissions reduction measures and further ways to increase the production of zero-carbon electricity by focusing on achievable solutions.

We also need better methodologies to measure both increases and reductions in emissions generated by AI. This fourth pole should include a combination of private sector-led action, intergovernmental initiatives, and public-private research efforts.

Managing significant increases in greenhouse gas emissions that could undermine climate goals, as the need for AI and increased electricity production is likely to take hold and accelerate further beyond current projections. is important. Governments, businesses, and others must incorporate the need for net-zero AI emissions into discussions about addressing the impacts of AI.

This oped was first published on The Hill

philippe benoit I am the managing director of 2050 Global Infrastructure Advisory Services. Previously, he held energy sector management positions at the International Energy Agency and the World Bank.most recently served as an adjunct senior fellow at Columbia University's Center on Global Energy Policy.

© Inter Press Service (2024) — All rights reservedSource: Interpress Service

Where next?

advertisement