By Alan AtKisson, Executive Secretary & CEO
AI – one of two “great disruptors” | How AI can accelerate IWRM | AI is not a replacement for new data, collective decisions – or muddy boots | Using AI to support better human dialogue | AI is here to stay – and we need to use it
If you are concerned about adapting to climate change – as you should be – then water is your principal worry. Suddenly there seems to be too much, or too little, or sometimes both, at different times. And the pattern is just too unpredictable. The timing of this relatively sudden change could not be worse.
In the past couple of decades, the world got steadily better at planning and managing water, for people, agriculture, cities, industry, and of course nature itself. But that progress needs to accelerate. Lack of good, integrated implementation may also lead to conflict between different types of water users, just when conflict in general is heating up in our world – and just when climate change is playing havoc with our carefully constructed plans for managing water-related risks and resources.
AI – one of two “great disruptors”
Meanwhile, if you are concerned about the impact of technological development – regardless of whether you are a utopian optimist or a dystopian pessimist – you are probably thinking about artificial intelligence. Some people think AI is the dawn of a new era of prosperity, some that it is the end of humanity. I personally think it is something in between: a new tool at our disposal, bringing with it enormous possibilities, and equally enormous risks.
I talk about AI routinely these days, whenever I am talking to groups about the future of Integrated Water Resources Management (IWRM). The general reactions always range from excitement to fear (or both), and the reflections on how we should use it stretch from enthusiasm to indifference to active hostility.
And yet, AI is here. Just as the Internet arrived in the 1990s to thoroughly transform our lives and societies, whether we wanted them changed or not, AI is about to do the same. And so is climate change.
That is why I have been thinking about the links between these two “disruptors” of global civilization, climate change and AI. For me, they go together, in a very specific way.
The world is changing extremely fast. We, in turn, must think faster about how to tackle the climate adaptation challenge and manage our water more efficiently and more effectively. And we also need to turn those thoughts into actions, as quickly as we can.
Meanwhile, AI’s express purpose is to help us tackle tough informational challenges much more quickly and efficiently. Which is another way to say: to help us think faster.
That’s why I am convinced that AI can help us – though I am equally convinced that it cannot replace us, as I explain later. Let me give you two recent examples of how AI can help that came up recently here at GWP.
How AI can accelerate IWRM
Recently I was listening to research colleagues from the Dutch water management research institute, Deltares, who are actively exploring AI applications in Integrated Water Resource Management, IWRM. (I was taking part in GWP’s global Technical Committee “Online Dialog” series, which are convened by our “TEC” Chair, Jaehyang So.) To make good water management decisions, you need good models – computer programs that can generate scenarios about what might eventually happen in a water system if you do X, or Y, or nothing at all. To generate good models, you need good data. To get the historical data and understand the trends over time, you have to “mine” the old data – and “mining” sometimes means literally digging it out from a pile of paper notes and logbooks compiled by some long-retired water-system technician or engineer.
Then you need to get that data into a spreadsheet, so that you can start to analyze it, to understand what makes that water system tick. It is an enormously important process, because the quality of regional agricultural planning for the Mekong Delta (just to pick one example) might hang in the balance. And the work is enormously laborious and time consuming.
Or at least, it used to be. Suddenly, we have AI systems with the ability to read anything, even old, handwritten notes, in multiple languages, extract the data, and put it into an organized data set. What used to take half a year or more is done in less than a week. These new AI systems, said Deltares senior hydrologist Gualbert Oude Essink, can do this kind of data-mining work “1000x faster and 100x cheaper”. Which is good news, because there are mountains of extremely useful data – data that could help us improve our models, plan better for climate change, and ultimately save lives thanks to better flood preparation or agricultural irrigation planning – that need to be digitized. Data that would otherwise be at risk of getting thrown out or recycled, as people retire or pass away.
AI is not a replacement for new data, collective decisions – or muddy boots
The benefits do not stop there. Using other AI systems powered by such data, we can develop better scenarios, which can support better dialogues (based on politically neutral information), which can result in better agreements in politically charged transboundary water negotiations – and thereby reduce the risk of conflict. This is a vision of AI working for people, supporting better water management, ultimately improving the conditions for peace.
There you have two positive examples of how AI could work positively in a sustainable water management context. Are there potential downsides? Of course.
For one thing, AI may fool us into thinking we know everything we need to know. But eventually we will have “mined” all the data that already exists, and nothing replaces new, real, verified data –actual measurements that some person, probably monitoring some automated sensor, has determined to be reliable and fed into the relevant statistical database. That is why AI is not the answer to all water data challenges, not by a long shot. We still need monitoring equipment, and people who know how to set it up, maintain it, read and interpret the data from it.
Can AI help us get more data, faster? Yes. We are years if not decades away from fully mobile water-engineer robots who can get their own robot boots muddy. But we do have satellites, and remote sensing, and ever-better AI systems that can interpret what the space-based or atmosphere-based sensors are seeing (as Kenji Nagata, Senior Advisor on Water Resources at the Japanese International Cooperation Agency, JICA, told this same online seminar group).
How do we make sure that the way we are using AI to support decision-making is somehow sound? That is a much tougher question to answer, but a critical one, as Åse Johannessen, senior research at Deltares, reminded the GWP TEC online seminar at its outset. This question is also driving a GWP-led initiative to set up an “AI lab” focused on water governance issues. How we use AI technology in human decision-making problems – how we apply it in governance situations, reliably and ethically – is ultimately a question that only humans can answer, too. (Although I could not help asking ChatGPT to write me a quick plan for attacking that problem. Which it did, in about 20 seconds.)
There are some areas of traditional IWRM practice that I personally would not want AI to do for us – for example, bringing people together to learn about what’s happening in a watershed, sharing perspectives, seeking a pathway to consensus, creating solid working relationships along the way that ultimately lead to action and implementation.
Using AI to support better human dialogue
AI obviously cannot replace human dialogue and negotiation among stakeholders. But can it support even that process? Theoretically, yes. During the GWP TEC online dialog, I typed a question into the premium version of ChatGPT, asking it to develop an approach for creating an agent-based model – that is, a model that simulates the behavior of individual actors or stakeholders in a system – in which the simulated stakeholders had different perspectives and had to seek consensus on a sustainable way to manage their common water resource.
In about twenty seconds, I had the outline of a plan for developing such a simulation model, including how to match the specifics of it to a particular basin, and to that basin’s specific stakeholder groups. I even had the basic structure I needed for writing the relevant computer code. (Assuming I can write Python, which I can’t.)
Do I recommend going that far? Simulating real multistakeholder processes, real people meeting in real transboundary riparian system, with AI-powered computer models?
No, I don’t, any more than I recommend moving to Mars, or doing space-based geo-engineering, to solve our global warming problems here on Earth.
But I do believe we need to explore how AI can best help us do some of the work that will empower those real people to understand their water resources better, and to come to better decisions faster, with better data and deeper understanding of what climate change is already doing to their water systems, what it might do in the future, and what they can do about it.
AI is here to stay – and we need to use it
What AI can never replace, no matter how good it gets, is the process of coming together, talking, learning, building relationship and trust, making choices about different policy options, and negotiating around values-based tradeoffs – the very human process that lies at the heart of all large-scale water management processes (and many other sustainable development processes).
AI can help, and we need it to help. We need to quickly learn a lot more about how it can help – and maybe hurt – such processes. Because whether like it or not, AI is here, and it is going to shape the way we work with water in the coming generation. That is the reality.
At least, until the next technological revolution comes along.
P.S. Yes, in case you are wondering, I am aware that AI server farms currently consume enormous amounts of water. Perhaps that is one of the first water-management problems we should use AI to help us solve!