AI and the environment ------------------------ As the media tires of using Bitcoin mining as an environmental punching bag, their attention is shifting to AI. Already, we are hearing a lot of the same claims "it strains grids", "it puts up electricity prices", "it uses too much water", "it harms local communities". They've even got self-styled researcher Alex de Vries pivoting from talking about Bitcoin mining and the environment, to writing about AI and the environment. As with the Bitcoin mining claims, most of these claims are either untrue, or lack nuance, but that doesn't get in the way of a good "Industrial villian" vs "community victim" narrative that has sold stories in the popular consciousness for hundreds of years. Having been part of a small group who spent three years changing the environmental narrative on Bitcoin, I can see a few of the same themes playing out. Overall, I think the AI industry is dealing with this better than the Bitcoin mining industry did. I recently saw a presentation from Melanie Nakagawa, Chief Sustainability Officer at Microsoft, which is 75% on the right track to heading off a lot of the false claims, and legitimately highlighting the areas where AI datacenters make a positive contribution. (The image below is from that presentation). And ... There are some legitimate challenges that the AI-industry face which Bitcoin mining didn't. Inference-AI such as your Chat-GPT query 1. is not FLC (flexible load consumers like Bitcoin mining). 2. has usage patterns which are often spiky which is hard for grid operators to counterbalance. The research tells us that unlike FLC such as Bitcoin, many large AI datacenters lead to grid operators to have to upgrade their grid (Norris et al), and place more fossil fuel generation on it (Rhodes et al), whereas Bitcoin mining helps defer grid upgrade costs and helps renewable energy transition. So the approach of the AI industry must be part message, part tech: 1. Learn from the lessons of the Bitcoin mining's 3 year journey turning around the narrative (while most of what the AI industry is doing is good, there's a few messages and strategies that they are using which we also used before realizing they are ineffective) 2. Use Bitcoin mining as 1-2% of total power-consumption so you can flatten out the genuine issues around inflexible and spiky demand. Advantage: Bitcoin mining more than pays for itself as it can also monetize otherwise wasted energy the AI-industry had to forward purchase I wish my brothers and sisters in the AI sector well in the journey ahead!(Daniel Batten)