IBM and NASA envision Climate ChatGPT

Future AIs from NASA and IBM want to take advantage of the petabytes of data the US space agency produces each year. A real treasure to prevent certain disasters related to climate change, as well as create a kind of ChatGPT for climate science researchers.

While AI models like ChatGPT and Dall-E have been touted to “change the world,” other binary intelligences have more serious missions than creating pictures of alien cats. Co-developed by NASA and IBM, and today formalizing a partnership around a big challenge: helping to study climate change more effectively. And try to find suitable answers. At the heart of the process are two treasures: data and model. The information belongs to the American Space Agency. According to Rahul Ramachandran, a researcher at NASA’s Marshall Space Flight Center (Huntsville, Alabama): NASA currently manages a volume of 70 petabytes, and our projections expect about 250 PB to be managed in 2025. “. An enormous amount of data, potentially rich in information… if only we could take advantage of it!

Also read: Space nuke action to save trips to Mars… and beyond! (January 2023)

Here’s where an AI model you know (perhaps unknowingly) comes into its own: the so-called “basic” model. If you’ve played with ChatGPT, you’ve used the special AI derived from GPT-3. It’s the same kind of high-end model that IBM wants to develop. Because they are the only people who can cover such a huge amount of information. And above all, as their name suggests, it is these models that can serve as the “foundation” for more specialized artificial intelligence. Just as ChatGPT generates text and Dall-E images, the IBM model, developed in conjunction with NASA data, will be the basis for scientific artificial intelligence. In particular, they will be able to help researchers, activists and other emergency services fight climate change. And their results.

A foundation, uses

Future AIs built on the underlying model could, for example, predict the movement of fires. ©NASA

While AI training models are abstract, AI applications are easier to understand. when asked ” How to explain what this model can achieve to someone who has no idea what artificial intelligence is? “, Rahul Ramachandran responds immediately with a specific example. “ Imagine rescue teams analyzing the image of a country hit by massive floods. With highly trained AI, the AI ​​will be able to quickly identify which areas are most affected thanks to a corpus of past and present satellite images. And help teams deploy their efforts smarter and more efficiently “. All this is enough to make an inventory of the analysis of the images, even if communications on the ground are interrupted.

Also read: Why NASA and ESA are betting on RISC V for their future space chips (Sep 2022)

The recording of satellite images was not accidental. Initially, the IBM model will work on Harmonized Landsat Sentinel-2 (HLS) data and images. Using the great power of such a model – no need for pre-labeled data – IBM’s intelligence will learn to recognize our planet. (according to the press release) ” to identify changes in the geographic footprint of events such as natural disasters, cyclical agricultural productivity, and wildlife habitats “. Basically, to be a kind of big brain that can see and scan the entire surface of the Earth and can support researchers in the interpretation of certain evolutions.

ChatGPT for earth science researchers

In addition to the visual parts of images (in true or false colors depending on uses and measurement tools), images also include additional information (terrain, temperature, etc.) that AI can automatically correlate.  So create links or offers that people might not think of.  © NASA
In addition to the visual parts of images (in true or false colors depending on uses and measurement tools), images also include additional information (terrain, temperature, etc.) that AI can automatically correlate. So create links or offers that people won’t think of. © NASA

Another intelligence that IBM and NASA will develop is artificial intelligence trained on 300,000 scholarly articles specializing in earth sciences “. An artificial intelligence that will work a bit like ChatGPT: become an easily searchable body of knowledge. Based in part on IBM PrimeQA’s multilingual question-and-answer system, this module is designed to connect to NASA’s data systems. And be able to be queried by researchers at the output.

It’s enough to develop tools where researchers can query the AI ​​and easily find its data and sources. Scroll through files faster than simple classic search engines based on keywords. And above all, by combining elements of knowledge that only AI can achieve. It is clear whether the relevance of the results will meet the expectations of a specialized audience. In any case, misuse of language and other elements of misinformation have the merit of at least being absent. Working materials are limited to scientific articles.

Also read: ‘Nothing revolutionary about it’: French AI pioneer Yann LeCun not impressed with ChatGPT (Jan 2023)

If these projects bear fruit, other initiatives are already being considered “Includes building a basic model for weather and climate prediction using MERRA2, an atmospheric observational data set. » It makes sense that this project follows on from the previous ones, because in addition to large volumes of data, the complexity of climate models requires additional computing power on a completely different scale.

Read also : It took IBM 20 years to protect us from a quantum threat that doesn’t yet exist (July 2022)

A final point for fans of public funds, unlike OpenAI and other models where many of the underlying technologies are closed, here the entire chain of knowledge is open. From NASA data in the public domain (images, articles, etc.), to IBM models to be licensed open source, the entire project is under the umbrella of NASA’s Open Source Science initiative. The question is whether IBM and NASA plan to design a “Climate ChatGPT” available to everyone!

Leave a Reply

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