This is the first in a multi-part series examining how artificial intelligence may impact California water.
It can be hard to escape the hype over artificial intelligence (AI) right now, especially if you live in California. Is it going to transform society for the better or destroy the world as we know it? And, critical to those of us working on water issues, what do these advances mean for efficient water use—and our jobs? Through a series of blog posts, we will explore how California might leverage AI to better manage our water resources, while mitigating the risks of this rapidly evolving technology.
AI isn’t new to California water, but recent advances may be changing the game
AI refers to computers performing tasks that typically involve human intelligence, such as reasoning, problem solving, and learning. While AI has exploded in the last few years, some water managers have been using early AI since the 1980s.
One of the most popular types of AI for water is machine learning, in which models learn and adapt without explicit instructions. In California, the Department of Water Resources and the State Water Resources Control Board have applied machine learning models to a range of topics, from modeling salinity to predicting drought impacts. Environmental NGOs and even private startups have jumped in as well, developing models to identify sufficient flows for ecosystems and inform water allocation. These applications likely have just scratched the surface: numerous academic articles propose more ways machine learning models can support water management in California.
Newer types of AI—particularly large language models that power the AI chatbots like ChatGPT—are making many tasks in the water sector faster, more efficient, and more accessible to those without specialized expertise. For example, these tools can rapidly summarize information (including legal documents), write computer code, and retrieve, analyze, and visualize data. California water utilities are using machine learning models for a variety of applications, including detecting leaks and increasing operations efficiency. There are countless other applications for California water, and many are still being discovered.
Today’s water data systems are often slow and clunky; some hail AI’s ability to rapidly integrate and analyze large datasets as transformative, potentially enabling a new field of real-time hydrology. This could help regulators, operators, managers, and water users make better decisions based on the most up-to-date information about current or forecasted conditions.
Risks of AI need to be monitored and mitigated
While these new tools hold promise for advancing California water management, conversations are just beginning about AI’s opportunities and risks.
First, AI is only as good as the data that it uses, and poor-quality data can lead to inaccurate or biased results. For managing water resources, long-term environmental datasets of temperature, streamflow, and rainfall are essential to getting good predictions. (As they say in computer programming, “Garbage in, garbage out.”) Recent cuts to federal programs for monitoring and data collection could harm the efficacy and accuracy of AI’s environmental predictions, including in California. There are data privacy concerns if water sector workers use AI on the job without proper precautions.
AI also can “hallucinate,” such as citing publications that don’t exist. And perhaps the most thorny challenge is that no one knows how AI reaches its answers, including those who built the systems. Building the data infrastructure, quality control, and trust in AI outputs is critical, especially as its use in the water sector becomes more commonplace.
Environmental implications of data centers
Finally, water and energy use by data centers, which house the computers that run AI models, has become an issue of growing concern. In California, the governor recently vetoed a law requiring data centers to disclose water use (often used for cooling), while signing a law to study the impact of these centers’ electricity use.
Decisionmakers in California will increasingly be called upon to set policy shaping the future of AI and water, given its risks and environmental challenges (though that effort may be complicated by a recent executive order). Smart policies will be needed to ensure that AI has a positive impact on California water while minimizing the risks.
link