AI & Climate Change

The two trends which will shape our future:

• Climate change

• Digital transformation

Goal: Both should reinforce each other. Digital transformation should help Europe with its zero-carbon economy

Proplems: Making it inclusive and affordable, e.g.: solar pannels on a rooftop are expensive

Mitigation –> reduce GHG-emission

Adaptation –> build resilience to the consequences of climate change

The Kyoto Protocol (1997)

• Decrease emissions by 5% from 1990 levels

• Industrialized countries and economies should limit and reduce GHG emissions

Paris Agreement (2015)

• The rise of global temperature to should be well below 2 °C above pre-industrial levels, and preferably limit the increase to 1.5 °C

• Emissions should be reduced asap and reach net-zero by 2050

Conserving natural resources and protect global ecosystems to support health and wellbeing, now and in the future

Different entities (UN, WHO, etc.) set standards for specific environmental sustainablities, such as quality of water, wildlife habitants, etc.

Exponential growth and profit-oriented economizing drives planet earth to a break point

Population growths leads to an increase of farming which leads to an increase of GHG-emissions

Chances are high that we will not achieve the goals of the Paris Agreement

Generally our practices as individuals and the world as a whole are not sustainable

Research communities like Climate Change AI believe that tackling climate change strongly depends on the help of Tech/AI

AI can help mitigigating and adapting to climate change

AI can also increase GHG-emissions and energy consumption

Compute-Related Impacts

• Model development and deployment –> when creating and running a ML-model energy is consumed. Since more complex models are developped more energy will be aquired

• Energy consumption of AI-models is difficult to measure –> basical rule: more FLOPS more energy

• Decision: Accuracy vs. Environmental impact

• Source of energy matters (renewable or not)

• Huge Datacenters are needed to store information –> energy consumption

Immediate Application Impacts

• Advatages of AI:

  • Forecasting

  • Calculating

  • Higher efficiency

• Disadvantages

  • AI can also be used for things which are bad for the environment

  • Higher efficiency of oil companies

  • Managing livestock –> cattle farming

System-Level Impacts

• Advertisment algorithms –> depends on the use

• Apps which encourage us to deal climate-friendly with our environment

• AI may change our general livestyle –> understanding nature better or getting estranged

Impacts can be good or bad

A study showed that training AI models to do Natural Language Processing (NLP) can produce carbon dioxide around the equivalent of 5 times the lifetime emissions of the average American car.

Support interdisciplinary development of climate impact assessment

Support interdisciplinary research in energy-efficient AI tools

Support and fund Machine Learning research for climate-related applications

Encourage clean electricity mandates

Enforce regulatory approaches to constrain sector-specific GHG-Emissions