Digi2al Data Scientist, Itai, opens up about public perception of AI and the impact of new technologies on our climate.
Whenever new technology arrives, we have to ask how it might impact our climate. While not exactly the new kid on the block, AI has become an increasingly pervasive topic in modern society, with ChatGPT taking centre stage. AI is often discussed in a personified, almost mystical way, perhaps because the mechanisms that constitute AI systems are highly technical and therefore opaque to most people (and many researchers, by the way). An interesting example of this is the now commonplace description of recommendation systems on social media and sites like YouTube as ‘The Algorithm(s)’. This is indicative of a general sense of unease about AI and how it interfaces with society.
While most public discussion about AI centres around its likely impact on people and society, there is far less discussion about the externalities of creating AI models in the first place, and the implications for our climate. When AI models are trained, especially models used in generative tools like ChatGPT or DALL-E, a huge amount of computation is needed. Intensive computation of this sort uses large amounts of energy, which, as with many other industrial processes in our economy, comes in part from burning fossil fuels. Wells Fargo predicts that, by 2026, AI power demand in the US could increase by up to 550% relative to the 2024 figure and 1150% by 2030. This means that AI would use approximately 16% of the total current energy consumption of the US. The bulk of current consumption is a result of training and retraining models, but inference usage is likely to increase considerably as AI tools become more common.
As AI becomes an ever bigger part of how the world transacts information, the effects of climate change on almost every aspect of society are likely to become more pronounced. The transition to greener energy sources is already one of the most essential issues of our time – and this only adds to it. It also means that making intelligent decisions about energy consumption and generation in the near term is critical. Many players in the AI space are well aware of this, and most of the large ones have signed net-zero pledges with various goals and timelines. The Chairman and CEO of Microsoft, Satya Nadella, recently announced that Microsoft “is looking for a Principal Program Manager, Nuclear Technology, who will be responsible for maturing and implementing a global Small Modular Reactor (SMR) and microreactor energy strategy.” If the projections discussed above are even remotely accurate, AI usage and development will become an increasingly important part of the overall conversation around how to combat climate change. Paradoxically, AI also has the potential to increase the effectiveness of current energy monitoring and management systems, so it will likely be discussed both positively and negatively in the years to come.