The previous LOTRW post noted the large size of the Conference on Artificial Intelligence for Environmental Science relative to other elements of the AMS Annual Meeting. The continuing growth of the AI conference portends a hugely expanded, positive future for AMS meetings and for our community—not for a few years, but for decades to come.
The reason? Powerful, enduring incentives for both the AI and AMS communities to collaborate, with multiple synergies promising big benefits for the world they serve. The (broad-brush) picture:
What’s in it for AI. AI is taking the world by storm. The community has two grand ambitions: to increase the breadth and depth of its problem-solving capacity, initially through the extension to artificial general intelligence and ultimately to artificial superintelligence; and to accelerate and spread the applications of that expanding capacity across every human endeavor. This combined effort and its fruits will shape the rest of the twenty-first century.
Weather forecasting,1 in and of itself, has long been an iconic challenge for AI’s parent IT world. When digital computers were first developed in the 1950s (and filled large rooms), weather forecasting was one of the first applications. The serial numbers of computers and supercomputers procured by the federal government for weather prediction were often in the single digits. Of course, advanced computation has since become more routine in all fields of science. Commercial applications and needs for supercomputing power now compete with the weather forecasting priority. With cloud computing, procurement is less of an event and more of an amorphous thing.
Nevertheless, weather forecasting remains a priority today, and an opportunity and sine qua non for AI. AI has shown forecasting skill: both as a stand-alone and as an adjunct to NWP and human forecasting. But far greater opportunity for societal benefit lies in AI’s contribution to the end use of such forecasts. AI is being applied to directly weather-dependent activity ranging from agriculture, energy demand, and energy production (e.g., from renewable sources such as wind and solar) to public health, water resources, waste disposal, and more. AI’s utility will depend in large part on how cleverly and fully it can integrate weather information into its inputs to everyday sector decisions and actions. Moreover, today’s highly urbanized, interconnected society relies on critical infrastructure’s resilience—its ability to remain uninterruptible even during flooding, hurricanes, tornadoes, and severe winter storms and their aftermath. In that ultimate sense, all of human activity, everywhere, is weather-sensitive at one time or another. Of note: The weather application, with its emphasis on predicting the future versus merely providing a more detailed characterization of the past, represents a significant next rung on that ladder to artificial general intelligence.
(BTW, the AI community itself has needs for better weather information to support its own work. Just one example: AI’s insatiable appetite for electricity and its growing dependence on weather-sensitive renewables and the integrity of power grids mean AI’s interest in weather and renewable energy has become and will remain especially keen.)
To meet societal/customer needs in this space, AI will have to draw on subject-matter expertise from the AMS and related communities. Given the time frames involved, it needs that expertise now to meet urgent coming demands. AMS offers a special opportunity for AI in that the community comprises and convenes researchers in large numbers—but not just researchers. For example, AMS meetings also bring together weather service providers, equipment manufacturers, broadcast meteorologists, educators, and policymakers. What’s more, AMS offers a path for greatly expanding the participation of the end-user community—and on a short time frame.
One special AI need that requires meteorological expertise and promises to be labor-intensive is the need for evaluation. End users require a clear understanding of AI’s weather-forecasting abilities and limitations at each stage of its development. They need to know when AI can be relied upon to meet critical user-sector needs, and when its outputs will have to be supplemented by other means. It’s evident that this evaluation is already lagging AI’s implementation, and that catchup will always remain a desired but elusive goal. Without it, AI’s potential for weather-sensitive sectors of the economy will remain suspect. AMS can help AI develop improved (faster, cheaper, more astute) means of evaluation.
Drawing on AMS-type expertise in the short-run will require that the AI community grow a bigger presence at AMS-type meetings, as well as incorporate some level of meteorological presence at AI meetings. AI is already paying attention to the pipeline of K–12 and higher education providing the needed pool of professionals. AMS can help in this task; it offers an extended educational network, as well as a local- and state-level footprint through its chapters.
AI can see incentives and clear benefits to its active participation across the AMS network of volunteer leaders and committee members, to engaging early-career professionals at the meetings, to support of students through scholarships, and more.
AI has other needs beyond such immediate, concrete concerns. Prominent among these are reputational/branding opportunities. Is AI an unalloyed positive, or a net positive, or a net threat with respect to AI’s potential benefits and threats to job markets, the economy, and even to ethical and moral issues? On this question the jury is still out in both the court of public opinion and at corporate and government policy levels. By working to improve public safety, and putting emphasis and effort on environmental challenges, the AI community can do much to improve societal outcomes and in that way burnish its image.
What’s in it for the AMS community. The AMS advances the atmospheric and related sciences, technologies, applications, and services for the benefit of society. Increasingly, progress toward these goals will be enhanced or limited depending upon meteorology’s ability to harness AI to improve forecast skill and its wide application, and by society’s effective uptake of advances in environmental forecasts.
For some time, such AMS ambitions seemed daunting, even elusive, given the accelerating pace of scientific and technological advance and accompanying social change sweeping the world as a whole. And that was before the cataclysmic events of the past year here in the United States: precipitous, wholesale layoffs of federal scientists; assaults on the budgets and independence of research universities; and the on-again, off-again character of federal policies and funding levels since.
AI offers hope with respect to both challenges. Over the past two years, even the nascent abilities of AI-enabled forecasting have proven game-changing. In response, commercial weather forecast services are proliferating. Dozens of startups are entering a space previously confined to a handful of national and international numerical weather prediction–based forecast centers. Meteorology has become much more innovative.
And in January of 2025, China released its open-weights “reasoning” AI model, DeepSeek R1, focusing minds. Suddenly the U.S. lead in artificial intelligence seemed fragile or even somewhat illusory—especially when it came to the broad application of AI. With that recognition came fears that AI supremacy is an existential matter of national security—that the future might belong to a single AI winner, with the rest of the world reduced to client states, and that China was vaulting into the lead, especially when it came to applications. The last time the world saw anything like this was the 1957 Sputnik moment in the Cold War. Back then, a sudden U.S. awareness of Russia’s lead in space technology prompted a U.S. push in science education and research to catch up. (And the U.S. did.)
Something similar is happening in the AI world today. The current administration emphasis is not so much in government initiatives per se, but in the private sector, and not so much in research and development per se, but in infrastructure. The Stargate project, calling for a half-trillion-dollar investment by OpenAi, SoftBank, Oracle, and MGX, is a centerpiece. Attention to the energy infrastructure needed to support AI is another thrust. In November, the Department of Energy announced a $300M Genesis Mission to “accelerate science through artificial intelligence.” Similar efforts appear elsewhere across federal R&D agencies. Administration documents also show awareness of the need for a well-educated and equipped workforce.
Unsurprisingly, the picture is fragmented. The larger initiatives have been hastily formulated and need fleshing out. The myriad lesser initiatives are disjointed and have been awkwardly inserted into agencies that are at the same time being dismantled and cut back. Immediate help for environmental research, though present, is limited.
Nevertheless, the post-Sputnik emergence and evolution of NASA provides clues for what to expect and reasons to be encouraged. The agency was formed from its predecessor NACA within months after the Sputnik launch. But priority for the first few years was on human spaceflight. Science missions began only a few years later. In the same way today, given U.S. concerns about a perceived Chinese lead in the applications of AI, and the economic and military implications of that lead, applications, even extending to environmental research, seem destined to start slow but survive and grow more robust with the passage of time.
It benefits the AMS to build its relevance to these thrusts. Though they represent an imbalance in what the field of meteorology might see as the most urgent priorities, they fund work that the community wants to do in any case and enable the community to build capacity and skills that are needed if meteorology is to contribute to the world of tomorrow. Importantly, high-level policy support for such AI-related work is most likely bipartisan. It will not dissipate with any change in the political winds.
What the world stands to gain.
Every human endeavor—the future of life on Earth—depends on the Earth’s natural resource bounty; on our ability to navigate Earth’s hazards; and not least, on our respect for Earth’s fragility. These three relationships have been radically reworked as human numbers and economic activity have increased over the past century or so. We are transitioning from a wild planet to a largely domesticated one. Our only future alternatives lie somewhere on the spectrum between wise, responsible management and thoughtless, destructive mismanagement. AI has already begun to accelerate this profound switch. Where the world ends up on this spectrum depends heavily on the effective collaboration between the AI and AMS-like communities, beginning now, and enduring for decades. Notably, success in this environmental arena could inspire similar positive outcomes in other areas of application. The world wants the AI–AMS collaboration to succeed. Let’s none of us neglect our part.
- And environmental forecasting more broadly. ↩︎
