AI benefits or beats the green transition path?

Meeting decarbonization goals in time to avoid the worst impacts of climate change will require an unprecedented amount of coordination between economic sectors in record time, which will be nearly impossible to carry out smoothly without machine learning. However, as the uptake of artificial intelligence expands, so too does its already enormous carbon footprint, raising the question of whether the use of AI in the decarbonization movement is essential or counterproductive. It’s a cath 22 situation.

The global green energy transition will require a scale and speed of systems transformation never before seen in history. In order to go as smoothly as possible, this process will rely heavily upon massive, complex, and nuanced computing power. Enter Artificial Intelligence. Around the world, AI is already playing a major role in renewable energy forecasting, smart grids, coordination of energy demand and distribution, maximizing efficiency of power production, and research and development of new materials – and that’s just the beginning.

The role of AI in the energy industry is about to take off thanks to a rapidly changing power sector, the scale of the transformation and the investments needed to make it happen, and an increasingly complex system of grid distribution and decentralization. Due to all of these rapid and sweeping systems-level changes, AI will be integral in ensuring maximum efficiency within decarbonization initiatives. Reaching net-zero greenhouse gas emissions in the energy sector alone will require infrastructure investments costing between $92 trillion and $173 trillion by 2050, according to estimates. AI has a massive role to play here, as “even small gains in flexibility, efficiency or capacity in clean energy and low-carbon industry can therefore lead to trillions in value and savings.”

The irony: all of this is going to require enormous amounts of computing power, which means enormous amounts of energy. Already, the carbon footprint of Artificial Intelligence is almost as large as that of Bitcoin – which is to say, equivalent to that of entire developed nations. “Currently, the entire IT industry is responsible for around 2 percent of global CO2 emissions,” Science Alert recently reported. What’s more, consulting firm Gartner projects that if business continues as usual, the AI sector alone will consume 3.5 percent of global electricity by 2030.

It has been estimated that training GPT-3, the predecessor of ChatGPT, required approximately 1,287-megawatt hours of electricity and 10,000 computer chips – that’s enough to power about 121 homes in the United States for an entire year. It’s also enough energy to produce around 550 tonnes of carbon dioxide. Indeed, it’s estimated that Open.AI, the creators of ChatGPT, spend about US$700,000 per day on computing costs alone to run its chatbot service for its 100 million users around the globe.

It’s very important to be careful with the use of AI to make sure that it does more good than harm – and that’s just in reference to its environmental impacts, leaving all of the other ethical and moral quandaries aside. The first major consideration when deciding how or whether to use AI in a renewable energy application is to determine whether it’s strictly necessary. Often, the use of AI can be more seductive than pragmatic. If it is necessary, engineers can next consider whether the energy used for the training is responsibly sourced, whether workloads are designed for maximum energy efficiently, and calculate and consider embedded emissions.