Advancing AI for Weather Predictions
NASA and IBM is developing an LLM with AI capabilities intended for meteorology and climatology purposes that should be much more effective than existing solutions. The second project in this area.
At the moment, there are existing AI models such as GraphCast and Fourcastnet that outperform meteorological models in generating weather’s forecast faster. Nonetheless, IBM insists that they are not the genuine LLMs but only AI emulators. In contrast to AI emulators, LLMs constitute basic tools that drive generative AI apps like GPT-4 for ChatGPT. This is a more powerful and fundamental way of doing things because even though the emulated AI can generate outlooks from the instruct training data they fail to encapsulate the mechanics associated with weather prediction as the two parties argue.
The objectives are different between NASA/IBM LLC and other existing models. It is aimed at wider distribution, faster inferences, and more comprehensive data instead. Nevertheless, the main objective is to improve the forecasting accuracy within different climate applications. The envisioned capabilities of the model include:
Forecasting Weather Events: As in other models, it must forecast weather incidents such as rainfall, snowfall, hails, and fierce wind.
High-Resolution Inference from Low-Resolution Data: That is why the model must utilize low-quality images, for example, from satellites, in order to generate an estimation of high quality parameters, including temperature and air pressure.
Identification of Conditions Favorable to Wildfires and Other Events: These factors include the drought, temperature and other conditions that can lead to more fires.
NASA worked with IBM in other special LLM in May. A geospatial model for monitoring and mapping tree planting within Kenya’s water retaining catchments. It seeks to grow trees and solve the problem of water shortage. The model is also used in studying urban heat islands in the UAE.
Commenting on these latest AI initiatives, Alessandro Curioni, IBM, Fellow and Vice President, Accelerated Discovery at IBM stated, “climate change, in itself, is a real and urgent problem for which we must seek new solutions using the fastest methods available including those with today
The future of AI and fighting global warming.
This is what he concluded; “Simple AIs based on geospatial data could provide an unprecedented tool that would help understanding, anticipating and dealing with myriad planetary changes due to global warming at a scale and with a speed like never before observed”. This could turn
The system will be trained and validated by teaming up experts of IBM with NASA professionals. More information about this interesting technology is to be provided whenever it becomes accessible.