NASA and IBM have collaborated to develop an unique use of AI technology on Earth observation data from NASA satellites.
The two organisations have stated their plan to develop several programmes to harvest Earth observation data for information about climate change. One project, for example, will use NASA’s Harmonized Landsat Sentinel-2 dataset, a record of land cover and land use changes obtained by Earth-orbiting satellites, to train an IBM geospatial intelligence foundation model. The AI will sift through terabytes of satellite data to evaluate how climate change and natural catastrophes affect agricultural output and wildlife habitats.
Meanwhile, another project will most likely include the creation of a large language model from Earth science literature. IBM developed a natural language processing model based on around 300,000 Earth science journal articles to aid in the categorization of the literature and the identification of new knowledge. Furthermore, the researchers have said that both models would be publicly available to everyone in the scientific community.
What are the plans?
Foundation models are AI models that may be used for a variety of tasks, are trained on large amounts of unlabeled data, and can transfer knowledge from one context to another. These models have rapidly extended the field of natural language processing (NLP) technology over the last five years.
IBM is in the forefront of using foundation models for applications not involving languages as a result.
How about the numbers?
Scientists are able to study and monitor our planet thanks to the enormous speeds and volumes of data that are being collected about it. To extract insight from these vast data resources, however, new and inventive methods are required. This endeavour intends to facilitate researchers’ analysis and information extrapolation from these enormous datasets. Furthermore, the speed with which these data can be located and analysed thanks to IBM’s foundation model technology may contribute to a better understanding of Earth and how it responds to climate-related issues in science.
IBM and NASA also seek to develop a number of cutting-edge technologies to gain knowledge from Earth observational data. In one experiment, an IBM geospatial intelligence foundation model will be trained using NASA’s Harmonized Landsat Sentinel-2 (HLS) dataset, a record of land cover and land use changes collected by Earth-orbiting satellites. Through the analysis of petabytes of satellite data, this foundation model technology will let academics to study the environmental processes that affect our world.
Artificial intelligence (AI) has advanced significantly over the past ten years, and it is hoped that its application to climate science could assist improve the accuracy of prospective climate estimates. Let’s examine a few of the current methods.
A group of American scientists showed in 2020 that time-series models and computer vision might be used to accurately model the dynamics of the Earth system.
The physics of clouds and rainfall processes will be modelled by artificial intelligence thanks to developments in this field, specialists from the Indian Institute of Tropical Meteorology in Pune, IIT Delhi, and others projected in 2021. This will reduce uncertainty in the existing systems.
Researchers from Google and Microsoft claim that, by utilising the high-quality data that is now available, AI has also been suggested for applications aimed at reducing the effects of climate change in addition to improving the representation of natural systems in climate models.
Other industries where AI is leading the way include carbon capture, building information systems, better transportation systems, and effective waste management, to mention a few.
According to New York University research, current deep learning algorithms have limitations, such as the inability to discriminate between cause and correlation. Furthermore, Moore’s law is expected to stop around 2025 as it encounters fundamental physical restrictions like as quantum tunnelling. Because of the increased demand for deep learning and other software paradigms, alternative hardware development is becoming necessary.
The future route
Another expected outcome of this collaboration between IBM and NASA is the creation of a searchable corpus of Earth science literature. In order to organise the literature and make it easier to find new knowledge, IBM developed an NLP model that was trained on around 300,000 journal articles in the Earth sciences. One of the largest AI workloads ever delivered on Red Hat’s OpenShift infrastructure is the fully trained model, which uses IBM’s open-source, multilingual question-answering system PrimeQA. The Earth science language paradigm may also be used by NASA’s operations for the administration and stewardship of scientific data.
In order to better comprehend climate change, IBM and NASA will also analyse and interpret Earth’s atmospheric data using MERRA-2. This collaboration is part of NASA’s ongoing Open-Source Science Initiative, which aims to foster an inclusive, transparent, and cooperative scientific community over the future ten years.