When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
As generative AI continues to gain ground globally, another type of AI — one that runs in the back-end of the industrial, scientific, healthcare, aerospace, and R&D industries — is also becoming more powerful.
These specialized AI models are lightweight, laser-focused, and incredibly fast and accurate, used to find new materials, predict weather, run complex biological, and physical simulations, and much more.
Along this line of development, NASA and IBM released a new AI model on Hugging Face — Prithvi WxC. The model is open-source — free for anyone to download, modify, and deploy.
NASA and IBM Roll Out the Largest Open-Source Geospatial Model
On September 23, 2024, NASA and IBM rolled out a new AI model for weather and climate prediction. The model — Prithvi Weather-Climate (Prithvi-WxC) on Hugging Face, is the largest open-source geospatial AI model to date.
To break down the news, we spoke with Johannes Schmude, Senior Research Scientist at IBM.
“Recent AI weather models are producing weather forecasts more rapidly and arguably as precisely as traditional meteorological models,” Schmude said.
Schmude explained that the industry has seen the emergence of large artificial intelligence models that focus on a fixed dataset and single use case; primarily forecasting.
“We have designed Prithvi WxC to go beyond such limitations so that it can be tuned to a variety of inputs and use cases. As a concrete example, the model can run both on the entire Earth as well as in a regional context.”
IBM and NASA’s foundation model will go beyond forecasting to solve larger climate challenges, along the way targeting improvements in precision, speed, and cost-effectiveness.
“With such flexibility on the technology side, Prithvi WxC is well-suited to help us understand meteorological phenomena such as hurricanes or atmospheric rivers, reason about future climate risk by increasing the resolution of climate models, and finally improve our understanding of imminent severe weather events.”
Schmude from IBM explained that AI forecast emulation — which uses historical data instead of conventional physics simulation calculations — has made amazing progress in the past years. However, unlike the new IBM-NASA AI, these models still run on ‘gridded data’, which is not the natural state of weather data.
Open-Source AI, Innovation, and Non-Linear Prediction
Sid Rao, CEO and co-founder of Positron Networks, an AI-driven platform for scientific computing and accelerated research, spoke with Techopedia about the challenges and benefits of the NASA-IBM AI.
“Training new foundation models, especially a multi-function model like what NASA and IBM built, is challenging,” Rao said. However, Rao recognized the potential of the technology.
“Specialized models like this are the next generation of foundational, non-linear prediction.
“Another reason to commend the NASA-IBM team is their commitment to open source — scientists should not have to require lawyers to understand whether they can build on the innovations of their predecessors,” Rao said.
Rao said that scientific achievement powered by public infrastructure should be released in the open, with permissive open-source licenses. Otherwise, the world is blocking national scientific leadership.
Rao added that AI models that make predictions based on dense historical data have already reached scientific domains. A great example of this is protein folding in computational biology.
Other examples Rao gave include fluid mechanics, solar weather, population patterns, economic theory, and flood frequency analysis.
“AI will disrupt pretty much any non-linear analytical task in science.”
AI Accuracy in Weather Prediction is Poised to Impact all Sectors
Dr. Max Li, an adjunct professor at Columbia University, and CEO of decentralized AI data provider Oort, spoke to Techpoedia about how the lack of reliable data sources creates gaps and deficits in predicting weather patterns.
“When experts can’t accurately predict even the track of a snow storm over a 24-hour period, it makes sense that it would be near-impossible to accurately predict climate change on a macro level over the course of decades.”
Dr. Li said that if NASA and IBM meet their promise, in a not-so-distant future, weather forecasters will not have to check different models to get an idea of what will happen, but rather just use one accurate, and relatable forecasting model.
“More accuracy in weather forecasting should lead to more reliable pricing for commodities including energy and crops. An ability to better account for variations in price for these materials could eventually lead to better hedging via futures and an overall across-the-board reduction in cost for raw and finished materials.”
Dr. Li added that all industries are poised to benefit from better weather forecasting — shipping, travel, agriculture, technology, hospitality, and even sports because extreme weather and climate change are increasingly impacting the ability of humans to function normally across the globe.
With 2.3 billion parameters, this foundation model is highly suitable for deployment on edge devices such as phones, sensors, and meters. Its flexibility opens a wide range of applications, making it an ideal choice for diverse use cases, Dr. Li explained.
The Bottom Line
The AI revolution is proving to go well beyond chatty GenAI bots. Scientists and developers around the world are teaming up to apply the power of AI to solve some of the most complex problems humankind has ever faced.
Predicting weather accurately, a vital necessity for society, especially under climate change impacts, has alluded the brightest minds in history.
NASA and IBM have taken the first step in using AI to solve critical global problems and encourage the scientific community to use its open-source model, democratizing access to the disrupting technology that is poised to change the world as we know it.
