Microsoft plans to reshape natural disaster modeling with cloud computing
Weather forecasting is known to be a complex field, but this task is becoming increasingly important for the everyday functioning of our planet. Climate change is intensifying the scale and impact of destructive natural disasters, from wildfires and typhoons to floods and cyclones.
Knowing exactly when and where a natural disaster will occur (or at least getting a warning several hours in advance) can radically change the outcome for affected populations. In this area, Microsoft sees not only an opportunity to do good, but also a profitable niche that can be used for its Azure cloud computing service.
As part of its AI for Earth program, the company has brought together a number of offerings into what it calls a "planetary computer." These offerings include APIs for identifying objects as well as plant and animal species. AI for Earth offers grants to scientists and others to use Azure in their research and modeling. The program joins the company's other cloud initiatives, such as AI for Health and AI for Accessibility.
If you are curious how the so-called planetary computer works and where the obstacles to more effective natural disaster modeling lie, you need to focus on all aspects of the disaster response system. The project's program director told TechCrunch that the project's ambitions remain as strong as ever.
“The goal is to have a planetary computer that helps everyone manage Earth's ecosystems, since this is the only way to be effective in the event of a disaster”
Although the program focuses on "mitigation, response and recovery," the response stage is the most interesting, since decisions must be made as quickly as possible. Nuño noted that over the past few years AI has advanced extremely rapidly, especially in areas relevant to the environment. According to him, artificial intelligence does not need as much data as many people think.
Great progress has been made in retraining algorithms. Scientists and engineers are working hard to help people understand AI and how an extremely effective deep learning model can be built. One of the major challenges in applying AI to Earth systems is the number of disciplines that must be combined for modeling to succeed.
However, many disciplines remain isolated from one another. Moreover, there is no wider gap than the one between scientists and AI researchers. Nuño sees the program as an opportunity to continuously engage all parties in tackling some of the most complex problems facing the planet.
“There are different dynamics due to different incentives: in the scientific community there is an incentive to create more knowledge, but for model developers it is to quickly get good answers. How quickly should decisions be made under uncertainty?”
One way to close this gap is what Nuño called "upskilling," or offering scientists the chance to learn more about artificial intelligence. All of this is part of the same strategy: helping people carry out environmental analytics faster and better. Nothing is more difficult than geospatial analytics. Computers are good at one dimension but not very good at working with objects located next to each other.
Sanchez-Andrade Nuño originally trained in astrophysics but upskilled in GIS (geographic information systems). Efforts to upskill in advanced AI skills have decreased as libraries have expanded, general AI modules work well, and there are colossal educational materials available to help people understand them.
The AI for Earth program director said that people used to need a doctoral degree, while today it takes 10 lines of code. However, this growing power leads people to believe that AI can solve any planetary problem. It cannot, or, if you are more optimistic, it cannot yet.
“We are trying to reduce the hype around artificial intelligence. If you do not know what AI is, you will not believe what it is”
The program emphasizes explainability in many of its initiatives so that scientists and AI researchers can together understand the results produced by models. This mission is increasingly aligning with relevant government agencies. Recently, the AI for Earth program signed a partnership agreement with the U.S. Army Engineer Research and Development Center to improve a coastal monitoring system.
Ultimately, modeling is becoming more mature, although much work remains. "Right now I would say we are still at the mezzanine stage. Many processes require a lot of special handling," Nuño said. The good news is that more and more people are entering the field, trying to connect the dots and, in the process, improve disaster response capabilities around the world.
