Weather forecasting is a notoriously difficult field, but it is a business that is becoming increasingly vital to the day-to-day functioning of our planet. Climate change is intensifying the scale and devastation of natural disasters, from forest fires and typhoons to floods and cyclones. Knowing precisely when and where a disaster strikes (or even receiving just a few hours’ notice) can dramatically change the outcome for affected populations.
In this area, Microsoft sees both an opportunity to do good as well as a lucrative niche to secure for its Azure cloud computing service. Through its AI for Earth program, which we last covered when it launched in 2017, the company has compiled a series of offerings into what it calls a “planetary computer.” These offerings include APIs to identify objects as well as plant and animal species. AI for Earth provides grants for scientists and others to use Azure in their own research and modeling. The program joins other Microsoft cloud initiatives such as AI for Health and AI for Accessibility.
I have been focusing on all aspects of the disaster response stack over the past few months, so I was curious about how the so-called planetary computer worked and where the roadblocks were to better modeling them. natural disasters. Bruno SÃ¡nchez-Andrade NuÃ±o, program director of the project, told me that the ambitions of the project remain as strong as ever.
âThe goal is to have a planetary computer to help everyone manage the Earth’s ecosystems, because that’s the only way to be effective in a disaster,â he said. While the program focuses on âreduce, respond and recoverâ, it is the response step that becomes the most interesting as decisions need to be made as quickly as possible.
SÃ¡nchez-Andrade NuÃ±o noted that AI has progressed extremely rapidly over the past two years, especially in areas related to the environment. âAI doesn’t need as much data as a lot of people think,â he said. âThere has been a lot of progress with conversion algorithmsâ¦ [and] we do a lot of work to bring people in to understand AI and how you can create extremely effective deep learning [model]. “
A big challenge for applying AI to earth systems is simply the number of disciplines needed to come together for modeling to be successful. However, many disciplines remain distant from each other, and no gap is wider than between scientists and AI researchers. SÃ¡nchez-Andrade NuÃ±o sees an opportunity for the program to continually engage all parties in tackling some of the most difficult challenges facing the planet.
There are âdifferent dynamics due to different incentives – in the scientific community there is an incentive to create more knowledge, [but] for model makers, [they want to] create good answers quickly, âhe explained. “How do you take decisions quickly in the face of uncertainty? “
One way to close that gap is to have what he dubbed âskill enhancementâ or to provide more AI training for scientists. âIt’s all part of the same strategyâ¦ to help people do environmental scans faster and better,â he said. Nowhere is this more difficult than around geospatial analysis. âComputers are really good at one-dimensionalâ¦ they’re not very good at things that are close together. He noted that he had originally trained in astrophysics, but had perfected himself in GIS (geographic information systems).
The effort involved in improving advanced AI skills has diminished as libraries have grown and common AI models perform well and also have prodigious educational materials to understand them. âPreviously you needed a PhD, and today you need 10 lines of code,â he said.
This growing power, however, is leading people to believe that AI can solve all planetary problems. He can’t, or if you’re optimistic, can’t Again. âWe are trying to reduce the AI ââhype,â he said. “If you don’t know what AI is, you won’t trust what it is.” The program has emphasized explainability in many of its initiatives, so that scientists and AI researchers can together understand the results of the models.
This mission is increasingly aligned with relevant government agencies. Recently, the AI ââfor Earth program signed a partnership with the US Army Engineer Research and Development Center to improve the agency’s coastal surveillance system.
Ultimately, there is growing maturity in a lot of modeling even though there is more work to be done. âAt the moment, I would say that we are still in the mezzanine phase,â said SÃ¡nchez-Andrade NuÃ±o. âA lot of processes require more ad hoc processing than we should. The good news is that more and more people are going out on the ground and trying to make the connection and, in doing so, improve global disaster response capacities.