Green AI Comes With a Catch

Among the dystopian predictions of how artificial intelligence will ruin the world, it is sometimes worth remembering that AI also has the potential to save it.

In climate change and sustainability, for example, AI is being harnessed in a range of applications that can drive more efficient and fuel-saving shipping and aviation routes, predict weather and enable precision farming.

These applications are also attracting significant early-stage investment. Microsoft has committed USD50 million to its AI for Earth Initiative, giving users access to AI for projects with a climate change or sustainability focus.

At the core of AI for Earth is the "Planetary Computer," which combines a multi-petabyte catalog of global environmental data with APIs and a flexible scientific environment that "allows users to answer global questions about that data, and applications that put those answers in the hands of conservation stakeholders."

The initiative is not purely altruistic because Microsoft will get valuable intelligence from how these startups use the technology. Then there is the potential to uncover exciting new companies which Microsoft might acquire or invest in.

Mitigating earlier

At the core of most of these applications is data, much of it relating to climate.

“Climate data sets are enormous and take significant time to collect, analyze, and use to make informed decisions and enact actual policy change,” said Jim Bellingham, executive director of the John Hopkins Institute for Assured Autonomy, at a presentation earlier this year.

"Using AI to factor in elements of climate change that are constantly evolving helps us make more informed predictions about changes in the environment so that we can deploy mitigation efforts earlier."

Google is also active in this area through its startup accelerator and supports several early-stage companies using AI in climate applications.

One Google-supported startup is Eugenie.ai, a platform that aims to help manufacturers in the metal and mining industries improve their decarbonization results.

“Using AI to factor in elements of climate change that are constantly evolving helps us make more informed predictions”

Eugenie offers a software-as-a-service solution that leverages satellite data and gives clients a comprehensive overview of their operations and a basis to track and trace assets. With these insights, it is claimed that companies can reduce their emissions by up to 30%.

California startup Refiberd is another one of these companies focusing on reducing the enormous amount of textile waste from the fashion industry. Around 200 million pounds of textile waste is generated yearly, but less than 1% is recycled into new clothing.

The Refiberd solution uses a patent-pending AI system to sort textiles more accurately, increasing the percentage that can be recycled.

At the other end of the corporate scale, IBM’s new geospatial foundation model has the potential to track environmental changes in minute detail and help humanity adapt to the new landscape.

Built from IBM’s collaboration with NASA, the watsonx.ai model is designed to convert satellite data into high-resolution maps of floods, fires, and other landscape changes to reveal our planet’s past and hint at its future.

The model should be available to IBM clients this year through the IBM Environmental Intelligence Suite. Potential applications include estimating climate-related risks to crops and buildings, valuing and monitoring forests for carbon offset programs, and the development of predictive models which could help mitigate climate change.

Human choice

All of these examples underline the fact that AI itself is amoral and is simply a tool to be used.

It can be used for good, and it can be used for less ethical purposes and the choice is in the hands of the human who makes the decision on how it might be deployed.

So rather than worrying about AI, we should put more effort into the frameworks humans will use as they work with it.

At the same time as using AI for climate solutions, there is a catch.

While these AI applications might help reduce emissions once created, creating and training algorithms is an energy-intensive process.

AI has its own carbon footprint. Researchers at the University of Massachusetts estimated that the carbon footprint of training a single AI requires energy equivalent to 284 tons of carbon dioxide equivalent, or five times the lifetime emissions of an average car, including the manufacture of the vehicle.

It's quite a paradox. Deep learning has made the AI revolution possible, and while it can contribute to decarbonization, its very creation is carbon-creating.

All of which makes the push toward green IT even more critical. Green IT will deliver greener AI, which will not only change attitudes to AI technology but will have an even more positive impact on climate defense.

Lachlan Colquhoun is the Australia and New Zealand correspondent for CDOTrends and the NextGenConnectivity editor. He remains fascinated with how businesses reinvent themselves through digital technology to solve existing issues and change their business models. You can reach him at [email protected].

Image credit: iStockphoto/Nuthawut Somsuk