AI generated image showing a female robot avatar wearing headphones emerging from a lake in the rain at night

THE list of reasons we should all be scared of AI just grows and grows. Now, to inherent cultural bias, misuse and the capacity to annihilate humanity, we can add drought.

Recent research suggests that AI’s global water demand will be the equivalent of a country the size of Denmark by 2027.

And’ as the use of data-hungry generative AI grows, you can see the problem. Our  ‘vampiric’ overconsumption of water has already brought us to the edge of a global crisis.

Will AI’s thirst be its undoing?


AI for social good

We need to think about it, because artificial intelligence has powers for social good that are almost beyond imagining. Think of ‘moon shot’ challenges such as curing cancer or ending famine. AI could transform social outcomes at scale across all of the UN’s Sustainable Development Goals.

In fact, we haven’t even begun to scratch the surface of what it’s capable of. No wonder the fourth industrial revolution is sometimes called the start of the imagination age.

In highlighting the under-considered impact of AI on water resources, the researchers might have done us a favour. AI developers who are working hard to improve governance and transparency over other risks are still largely ignoring water. It’s a good moment to tackle that omission.


AI water use explained

AI’s water usage comes in two broad forms, the on-site cooling of data centres where AI servers generate massive heat, and the off-site water consumption used in energy generation to power those data centres.

Paranoia has taken root following the astonishing growth of generative AI. Although not definitively linked to a sudden huge leap in data centre water consumption, its widespread adoption in around 2022 curiously coincides with it.

That year, Microsoft reported a 34% increase in data centre water usage. To add to the furore, Google followed reports of its own 20% increase with an admission that increased use of AI is making ESG targets harder to hit.


Could AI be the solution?

Which makes you wonder: If AI is such a power for social good, why can’t it sort out its own water problem?

I’m glad you asked, because it can.

But first, remember that AI in itself is not the problem. Man’s lax attitude to his most precious resource pre-dates AI by decades, if not longer.

Years of under investment, coupled with climate change, population expansion and economic growth, has left our systems of water infrastructure creaking, old, and increasingly in the wrong place.

In fact, we mismanage our water to a mind-boggling degree. Around 30% is lost within networks before it even reaches the consumer. That’s a huge financial drag on water utilities, because it’s lost revenue they can’t re-invest in asset renewal.


Water AI needs to think bigger

With AI this is changing. Like other sectors, AI is transforming the efficiency and reliability of water utilities by augmenting humans’ ability to make sense of data and take the right decisions at the right time.

By rapidly, accurately and indefatigably executing operational tasks which would exhaust human capacity, AI is analysing real time data analysis from digital sensors so that issues are seen and fixed earlier. And, even better, turning this ‘sensorisation’ into digital twins – virtual data models offering up precision AI insight for nuanced human judgment at a systemic level.

So far so good, but the UC report is a wake-up call. With global water conflict on the rise and generative AI on the march, this approach is not going to be enough


Water is not a zero-sum game

One big problem is that water is not like other resources. It’s messy and unpredictable. An intervention in one place can have serious unexpected repercussions somewhere else.  Traditional approaches of negotiating water shares is flawed. Water is not a zero-sum game.

And it is not a problem society can expect water utilities to handle on their own. Many of the world’s most stressed water networks don’t have the luxury of digitisation. This limits their use of the most sophisticated AI tools, or they find that the data quality from older sensors is not good enough to inform accurate decisions.

To guarantee its own future, AI urgently needs to engineer its own adoption in the water industry, and provide a bias for collaborative action which draws multiple non-traditional partners into the challenge.


AI as the engine of water technology adoption

One sector to respond to the need for collaborative action, and the potential of AI to address it, is the data sector.

Microsoft leads the global charge. For years, it has worked towards a water positive goal of replenishing more water in stressed watersheds than its data centres pull out of them.

FIDO AI’s ability to identify, locate and, uniquely, size water network events like leaks, gives companies like Microsoft the potential to rapidly accelerate their progress.

By supporting our AI use within utilities to address leakage, Microsoft is driving immediate, local, and transparently quantifiable resilience to claim against its target.

And that’s not all. By demonstrating the volumetric benefit of any water tech, FIDO can attract new partners into the water sector to help get them adopted and scaled.


Catalytic communities led by AI

This model of AI-led water management which accelerates the adoption of what the UN calls ‘beyond business as usual’ approaches to water scarcity has now become the world’s first catalytic community for water in the Colorado river basin.

The potential is enormous.

As well as tying diverse stakeholders into commercial agreements which deliver individual benefits with a common goal, these communities inject new finance into the water industry and foster a mutual understanding of the complexities of water management.

Trained on global water data from multiple sources, FIDO AI knows which data it can rely on. Soon, with just a sample of new information, we will be able to plug regional data gaps and build a reliable network-wide baseline information. It’s only a small step from there to integrate other publicly available data sets and understand the impact of interventions across an entire watershed.

And as for generative AI, using LLM our askFIDO is already bridging the technology and skills gap at an engineering level. AskFIDO provides humanistic answers to questions about where to dig. Why not answer bigger questions to support discussions over water allocations, building projects or infrastructure development?


AI: water’s friend or foe?

If AI achieves anything, it’s in allowing us to extract truth from the certainty of data.

With AI we can plug gaps in our knowledge and skills, decipher myriad watershed interactions, actively collaborate across state, sector or interest boundaries, deliver a bias for action or even defuse conflict.

For the longest time, water has been considered a cheap (if not free) resource, a human right and fair game for industry to use and pollute. It’s only recently that we’re really beginning to understand that only one of those statements is true.

So, yes. Let’s be honest about AI’s problem with water, without forgetting its potential to solve it for all of us.

Victoria will be moderating a SWAN APAC panel on the implications of generative AI for utilities at Singapore International Water Week on Tuesday June 18, 2024.


Subscribe to Victoria’s monthly Linked In newsletter Losing It here and make sure you get her blogs first.