The artificial intelligence revolution in climate science

We have just witnessed the beginning of a paradigm shift in the geosciences. An article published in the magazine nature A study in July showed that a neural network (artificial intelligence) predicted the weather better than the European Center for Medium-Range Climate Forecast, which has the most advanced climate forecasting system in the world. And then, in November, Google's DeepMind advertisement Artificial intelligence for weather forecasting has generated more accurate forecasts.

Traditionally, the approach to climate prediction is to use observations taken at a point in time as initial conditions to create equations based on physical principles.

In contrast, AI will process data collected over time and then “learn” the dynamics that traditional equations should clearly describe. Both methods rely on supercomputers, but AI does not need formally developed theories.

Weather forecasts determine aircraft destinations and ship routes and help manage all types of civil and military risks caused by the changing environment. That matters. While these are relatively early days for AI applications in this area and there is still much room for development, both in this sector and others, AI-powered forecasting may replace the jobs of human professionals, since neural networks do not require knowledge of meteorology. Dynamics (authors of the article nature They are engineers without this kind of training.) But the effects don't stop there.

In his writings on the problem of statistical prediction in the 1950s, Norbert Wiener, the father of cybernetics, I noticed If we already know the history of a system that exhibits certain properties, adding knowledge to the equations that govern it will not necessarily improve its predictions.

See also  Medical graduates: competition and migration

Wiener was, at the time, addressing a largely theoretical issue, as limitations on observations, data, computer processing power, and other factors did not allow for much more. But his argument now goes to the heart of the problem and captures the larger implications of recent advances in artificial intelligence.

Space infrastructure

In just a few years, we have grown dramatically Our monitoring data from Earth. Between 1993 and 2003, 25 observation satellites were launched into space, but between 2014 and 2022 the number rose to 997, bringing the total fleet of Earth observation and other satellites in orbit to about 7,560.

With a large-scale space infrastructure transmitting data on almost everything from planetary growth, water vapor, utilities, height of forest cover, and measurements of the state of the atmosphere, we have entered a new phase. The golden age of Earth observation.

This growing archive of data describes almost everything nature does and what we do on Earth. Combined with new AI models and ever-expanding computational processing infrastructure, it could turn our understanding of the planet and our role in it upside down.

Let's think about climate change. For the past 40 years, humanity's response to the climate crisis has been guided by the Intergovernmental Panel on Climate Change, a scientific body divided by disciplines: The physical sciences use large-Earth models that have many things in common with those used to predict climate. While economists and geographers, separately, measure the impacts and emphasize the role that adaptation and mitigation policies play on our societies.

This division of labor – is reflected in Tripartite working groups IPCC – corresponds to the division of methodologies. While physics-based Earth systems models rely on first-principles equations, economists and impact modelers use a combination of empirical methods and irreducible theories.

See also  NASA Trains Puerto Rican Students Interested in Oceanography - Metro Puerto Rico

AI could disrupt all of this. Although they are unlikely to completely replace traditional climate models (our observational record is not long enough to provide a rich statistical picture of climate phenomena over the centuries), they do play a role. a big role In this field.

Practical use

To be more specific, what matters most to us is not how the climate system behaves, but how it affects the world in which humans and other creatures live. AI models, independent of any existing scientific theory or disciplinary paradigm, can do this Help us conclude And perhaps predict changes in natural biomass over time.

This, in turn, could improve the way we manage forests and agriculture, and build diagnostic tools and early warning systems for risks Fires also FloodsWe understand how Energy economy associated with these changes or expect their effects on The economy in general And even in Climate negotiations. All this in addition to how artificial intelligence works Speed ​​up the transition process Towards a low carbon economy.

Of course, artificial intelligence is no substitute for scientific understanding. Science must remain a fundamental human drive, where the value lies more in asking the right questions than in simply deriving the answer from the data.

Regardless, we should make the most of the cognitive shift initiated by AI: it can help us identify new observable phenomena that have not yet been detected by disciplinary lenses. It can help us manage natural-scale systems that are too complex to be subjected to theorizing. It is the ultimate exploratory tool for blurring boundaries between disciplines.

See also  UNLP opened a "historic" work for the Faculty of Exact Sciences

This change also represents a profound political challenge. The infrastructure that powers them – Earth observation satellites and computer processing – is increasingly under private sector control. the The largest satellite owner The Earth observing company is a company called Planet Labs.

Big tech companies (such as IBM, Nvidia, DeepMind, or Huawei, whose employees wrote the July article in… nature) is at the forefront of machine learning. With access to unparalleled capital and resources, they are companies that can easily outperform most government think tanks. Some may be very charitable, but they are not obligated to provide public goods or care about equitable access to infrastructure.

As we struggle to deal with the ramifications of the digital revolution and the natural environment that is changing before our eyes, artificial intelligence may be key to unraveling some of the complexities that our understanding has failed to capture. However, with research tools in private hands, policymakers will have to be vigilant in ensuring that these new tools provide public rather than just private goods, and that the questions posed to them produce responses that help shape countries' legitimate policy. Goals.

Giulio Boccaletti, Scientific Director of the Euro-Mediterranean Center on Climate Change, and author of the book Water: a biography.

© Project Syndicate 1995-2024

Leave a Reply

Your email address will not be published. Required fields are marked *