The AI Citizen and Climate Change

In Search for Data Science’s best Levers for Climate Action

While I was in London, taking part in a conversation on how technology leaders can help meet the UN Sustainable Development Goals, a very interesting workshop was being held in Long Beach: ‘Climate Change: How Can AI Help?’ I wish I had been able to attend. Headlining the workshop was a very detailed review paper on the topic, titled Tackling Climate Change with Machine Learning, which I highly recommend reading. It’s a great answer to questions that I, a machine learning scientist eager to learn how I can contribute to addressing this pressing issue, have been asking myself for a long time.

Figuring out what role one can play, both as a domain expert and as an individual citizen, in tackling the challenges associated with climate change is not simple.

The majority of problems that machine learning is well suited to answer revolve around prediction and optimization, which could presumably provide solutions to issues around improving energy efficiency. On the surface, better efficiency seems desirable, but it is often paradoxically a double-edged sword: since demand for energy is essentially infinite, more efficiency directly translates into lower energy costs, which increases the availability of cheaper energy, and leads to more demand. This cycle is known as Jevon’s paradox and largely holds true for resources with very elastic demand: planes for instance are more efficient than cars, which are more efficient than horses, but that doesn’t mean we use less jet fuel today than we used horse feed then. Efficiency often drives more consumption, which is not great when you’re making fossil fuels cheaper as a result. If you’re interested in the dynamics at play, and how complex thinking about energy can be, I recommend the (slightly dated but) very provocative book The Bottomless Well.

There are areas where more efficiency does mean a better advantage for renewables, and the workshop paper does a good job at highlighting those opportunities, whether it’s about better forecasting of energy demand or better material design. There is also a very large opportunity in bringing the best that data science has to offer to scientific fields that have not traditionally been data-driven and but instead rely predominantly on mathematical modeling. It’s always struck me how much of our entire technological progress is often bottlenecked by battery technology: Improving the energy density of storage by a few factors at equivalent cost would completely change the transportation landscape, mobile computing, and tilt the equation unequivocally in favor of renewables. The most daunting task, however, is not simply to enable more use of renewables and to put a stop to carbon emissions, but to find ways to bind the existing carbon in the air in a more efficient way than evolution has managed to come up with in the billions of years of tweaking photosynthesis. Surprisingly, attempts here are not without precedent.

Richard Curtis on leveraging tech entrepreneurship towards achieving the UN Sustainable Development Goals. Founders Forum London 2019.

It is also important for us, the machine learning community, to not become part of the problem in the first place. We collectively use a lot of energy. While I am extremely fortunate to be part of an organization which matches 100% of its energy use with renewables, most other infrastructure providers, even the big ones, struggle to get there, and it’s not for lack of trying and investing in it. Offsets are also not the complete answer, since they can only make a positive difference in a regime where carbon-based fuels are still an alternative to renewables, and may ultimately detract from the goal of eliminating them altogether. Nonetheless, today’s world is still one where these forms of energy can be traded against each other with varying degrees of utility.

Accounting for my personal impact as a citizen brings up a very different set of questions. My home city of San Francisco provides us the choice of using 100% renewable energy sources for electricity, which is fantastic. However I, like most of my neighbors, still rely on natural gas for heating. I’ve actually wondered how many pounds of carbon I would have to offset to be ‘neutral’ in terms of direct energy consumption. Thanks in part to my solar installation, I am left mostly with air travel to account for, which, sadly, dwarfs all other direct carbon emission sources for my household (Since last year, I purchase offsets for that footprint, too). Even though I try to reduce my own air travel, I remain ambivalent about the ‘travel shaming’ trends that are entering the zeitgeist on the basis of the carbon impact of that industry. Mark Twain’s famous quote, ‘Travel is fatal to prejudice, bigotry, and narrow-mindedness’ expresses a sentiment that I can directly relate to. I can’t imagine the person I would be today if not for easy access to a world so different from my rural France, and how foregoing that experience would have shaped my own thinking on global citizenship and impact on the planet at large.

If I wanted to account for all my indirect impact — meaning the environmental impact of the things that I purchase and use, it would seem, on the surface, to be difficult, particularly in the absence of the exact provenance and energy footprint of the things I consume. Interestingly, and controversially, there is a crude shortcut one can use to estimate it: to a first order approximation, your entire footprint is the number of kids you have, times a constant factor. Kids will have kids, and each direct descendent you have today compounds to a number of additional human beings on the planet. Each will individually yield a degree of carbon-generating consumption that’s largely independent of whatever action you take today, and that will have more to do with lifestyle choices, technology and economic context of future generations. That aggregate consumption, under reasonable assumptions, dwarfs your own. Hence the constant factor, what you do today to influence your indirect footprint has little bearing to the amortized decisions your descendants will make, unless you believe that sensibilizing your kids to the issue can have material trans-generational effects, and the discount factor to apply largely depends on how long you believe this crisis will take to unfold.

Hearing that your best lever to limit your environmental footprint may be to have fewer kids is yet another angle by which the issue of climate change can tickle the world’s religious and geopolitical neuroses. It is also a poor excuse to pass the buck to the next generation. There is broad consensus in particular that the best and most urgent levers to pull in the fight against climate change are political. Many of my Bay Area neighbors feel quite disenfranchised on this front, given the modest role we can hope to play on the national US stage. Perhaps having an environmentally-minded Oaklandite in the presidential race can change that. On the regulatory front, I am personally very eager to see how Canada’s foray into revenue-neutral pricing of carbon emissions unfolds, because we need more large-scale policy experiments like this one to bring about change.

Even though climate change is the issue of our times, I can’t help but wonder to what extent it is at risk of overshadowing the other environmental problems we’re facing today. Many of these issues are beholden to the evolution of the climate, but many others are at risk of neglect in the face of a worldwide, invisible crisis unfolding quickly. What happens when biodiversity and habitat preservation run counter to the sustainable development of renewable energy sources? Or when we collectively decide that preserving the world’s forests is no longer a meaningful enough lever in this fight that we should care? I fear that increasingly, anthropocentric responses to the climate crisis are going to force us into tradeoffs that run up against the broader goal of ecosystem preservation.

We are at risk of losing sight of what’s worth saving in the first place, and the solutions we come up with will force us into very difficult compromises about how we choose to shape the future of planet Earth. Nonetheless, it’s important to think about how one may best employ one’s strengths towards these important social goals, because that’s where we can likely find the most leverage to make a material difference in outcomes beyond merely making good personal choices as a citizen.

I am a Principal Scientist at Google, working on Machine Learning and Robotics.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store