How is AI helping us to battle climate change?

WorldSummitAI is a yearly conference on artificial intelligence that is organized in Amsterdam. Increasing vaccination rates and relaxation of covid measures allowed this event to take place as a partially physical event again this year. With a particular focus on AI and climate change, the T-minus30 event was organized as part of the larger event. As a researcher from the field of drinking water, I find these meetings interesting because they may be inspirational towards new applications and opportunities of AI in the water sector, in particular in the context of climate change, which is in my opinion the single greatest challenge that our sector (and humanity) has to deal with. And after all the virtual conferences of the previous 1.5 years, actually visiting one in person did make a nice change. But is that enough to get back to our pre-covid travel-intensive conference-hopping habits? In the following paragraphs, I would like to share some of my impressions and thoughts with you (note that this is not a comprehensive overview).

As part of the plenary program, there was a panel discussion (with only online participants) on the question whether there is an urgent need for ethics in AI. Responsible AI was already a spotlight topic in the 2019 edition of WorldSummitAI (see also my blog  describing my impressions of that event) as a precursor to the concept of ethical AI. But the use of AI in dealing with the pandemic that has held the world in its grip ever since just a few months after that event has made the urgency of this issue much clearer. The possibilities and fear that data collected in controlling the pandemic could be used to control or enforce non-related social behaviour is an example of that. Very interesting was the Chinese perspective on ethics in AI. This has, on the one hand, values (such as autonomy) that have originated from the west but have become almost universal. On the other hand, the perspective builds on a tradition that includes philosophical systems such as Confucianism and Daoism that lead it to put more focus on the group, flexibility and collaboration. A recent change of attitude with respect to ethics in AI was recognized by one of the panelists in the growing role of privacy and data ethics in legislation. The final panelist, from the Center for the Study of Existential Risk, touched upon the work of his institute on the question how humanity can survive under new conditions brought about by technological developments. The most relevant example for the present context is AI becoming smarter than humans. The discussion on ethics in AI possibly became a little more concrete in the following presentation by prof Emile Aarts. He described the ELSA labs that have been set up to include Ethical, Legal and Societal aspects in a novel human centered AI approach.

Slide from prof. Emile Aarts’ presentation.

So have we advanced in the past two years in terms of ethical AI? Not necessarily from a technical perspective (at least, judging from the discussion and presentations), but my impression is that awareness of the issue has grown significantly and that there are better legal frameworks in place now.

The first block of the climate-focused T-minus30 program, having a panel discussion and three presentations, focused on AI-supported disaster relief and sustainable food production, with a focus on proteins, plant-based and cultured (which also uses plant-based proteins). The role of AI in this latter topic may not be immediately obvious, but it is there. First, AI may help to enable the food production and supply system becoming more circular, by feeding data on consumption and demand up the chain. Also, AI may help in innovation, helping to better understand and predict biological processes, such that biology may become engineering (in the words of one of the panelists). As a researcher from the field of drinking water, I found this very recognizable – it is exactly that which we are aiming and working to apply AI to: better understanding and predictability of our systems in order to improve their efficiency and resilience, and increasingly also their circularity. Also, the assertion that a lot of time is still spent on curating data was very recognizable. This is a returning point of concern in our projects and of discussion in our research programs.

Prince Jaime de Bourbon de Parme, our national climate envoy, gave an overview of what to expect for the next climate summit COP26 in Glasgow next month (steps in the right direction, but not enough), and an overview of applications of AI to help mitigate or adapt to climate change that he came across in his work. These include predictive analysis for carbon management, process efficiency maximization, forest management and protection (including wildfire forecasting), and refugee predictions to anticipate the need for aid.

I think that these examples translate quite easily to water management and supply, in particular the third and fourths examples. The third can contribute to the monitoring of catchments and the prediction of runoff and resulting floods. The fourth example on humanitarian aid can be applied both on short time scales to predict water demand for immediate relief and on longer time scales in the context of climate-change-induced migrations that are expected to have significantly increased by mid-century.

Slide from Dr. Peter Lee’s presentation.

The final contribution that I would like to mention here was by Dr Peter Lee (Microsoft’s corporate vice president of Research and Incubations) on societal resilience. He identifies the new paradigm and research field of crisis response science – that is to say not the study of how we should respond to crises, but the way in which doing science has changed as a response to a large crisis in society: the covid pandemic has shown a very rapid and effective response from the scientific community to identify the threat, understand its propagation, and propose and develop mitigative and curative measures. The crisis has also illustrated how the public debate and even politics may enter into and influence crisis response science. Society has shown some degree of resilience in dealing with the pandemic. The three pivotal elements of societal resilience (or at least the need for them) that came to the surface are: 1) a ready pool of applicable technologies, 2) support for collaborations across boundaries, and 3) community leaders as first class partners.

Research is being done on water (supply) system resilience in our institute, and has been for half a decade. Additional steps should include embedding this in the concept of societal resilience; the recently started Smart Water Futures project (, in which KWR is a leading partner, is looking at the resilience question from multiple perspectives in the larger societal context, so that’s a great opportunity.

So how is AI helping us to battle climate change? Considering what I have seen and heard during the first day of the conference, AI can definitely contribute to making individual systems (such as food or water supply) more efficient and more circular, in this way reducing the pressure on System Earth. AI can also help to better understand what is going on in the world; how the planet is responding to the stresses we expose it to. But what I also observed is some degree of myopia, with a blind techno-optimism that fails to consider the question whether the technological solutions provided outweigh the damage that they, or the practice or system that they support or optimize, may cause (in terms of energy and resources required, health, etc.). AI may help to improve the efficiency of cultured meat production (i.e. in a vessel rather than an animal), but this steps aside of the question whether directly consuming plant-based proteins might be the most efficient and sustainable route. Or AI may be very useful in supporting the electrification of on-ground airport operations, but the question of how and on what scale sustainable aviation in the current paradigm (jet planes with biofuels or synthetic fuels) is even possible has not been answered yet. These AI-based technological solutions are great, but their implementation needs to be viewed in the bigger picture, with a systems thinking hat on.

And then to get back to the question of virtual vs. real conferences. I always find it difficult to resist the temptation of doing something else during a virtual conference. Actually being in a room with a speaker and an audience took away that temptation, making the transfer of knowledge and ideas more effective for me. I’m not sure I would have been able to write the same blog based on virtual presence with currently used technology. However, as mankind desperately needs to cut down carbon emissions by 50% or more in this decade, I also feel that we cannot get back in the mode of frequent long-distance travelling for conferences. The conference-crowd needs to cut down their emissions by 50% (at least) as well – there is no room for exceptions. And since the aviation industry is not going to manage doing that with technological solutions, this reduction will have to come from travelling less far and/or less often. We need more experimentation and experience with alternative, more immersive forms of virtual conferences that offer complete interactivity with other conference participants so that it will take away the temptation of doing something else.