REDMOND, Wash. — Lucas Joppa agrees we’re living in the Information Age. But he wishes that the present tech era wasn’t so navel gazingly focused on Homo sapiens.
“I want an Information Age that encapsulates all information about life on Earth,” said Joppa, who is Microsoft’s first chief environmental scientist — and likely the first chief of this kind anywhere in the tech sector.
“We’ve allowed ourselves to exist in a world where we’re completely flying blind to the rest of the life on Earth,” Joppa said. “We do that at our own peril, and it exhibits an exceptional lack of wonder about where we are and who we are and why we’re here.”
Joppa is leading Microsoft’s new AI for Earth initiative, a five-year, $50 million grant-making project to unravel some of the world’s non-human mysteries. It’s helping environmental groups and researchers use artificial intelligence, machine learning and a variety of the company’s cloud-based tools to further their eco-causes.
Since its December launch at the Paris climate event, the initiative has awarded 112 grants for projects from 27 countries and 25 U.S. states. The AI for Earth project has four focus areas: climate, agriculture, water and biodiversity. Some of the initial projects target wildlife population surveys through image- and sound-recognition tools. The group has also produced a tool that generates detailed maps showing U.S. land use.
The project is unusual for working across the tech giant’s numerous departments and products, and fits into the approach of company-wide collaboration championed by CEO Satya Nadella and company President Brad Smith. Earlier this month, Microsoft announced the similar AI for Accessibility initiative, with a five-year, $25 million budget and a focus on helping people with disabilities.
Last week Microsoft hosted some of its AI for Earth grantees at a conference on its Redmond campus. They provided training for some of the cutting-edge tools and services available through Azure, its division of cloud-based offerings.
While the partnerships benefit Microsoft through positive publicity and by cultivating new customers in the environmental sector, grantees say there’s real potential for technology-fueled breakthroughs in conservation science.
Seattle’s Snow Leopard Trust is working with Microsoft to develop tools for estimating the population size and location of the elusive, Central Asian cats.
Using camera traps, the nonprofit has captured roughly 1.3 million images from the leopards’ rocky, mountainous habitat and aims to gather half-a-million images or more each year. But the cameras are triggered by all sorts of movement and other animals, so only about 5 percent of those pictures actually contain one of the well-camouflaged cats.
It’s one of the “largest data sets of goats and grass blowing in the wind,” joked Mark Hamilton, a Microsoft software engineer assisting the nonprofit.
Sifting through the existing images would take roughly 19,500 hours of human labor. So the trust and Microsoft have employed deep learning, a technique that uses “brainlike algorithms” that learn to sort through the photos automatically. Engineers have created a scalable image recognition program that is nearly 95 percent accurate in identifying snow leopards.
Hamilton additionally created a live dashboard that highlights snow leopard hot spots where the cats particularly like hanging out.
The snow leopards’ habitat spans 12 countries and an area that’s more than twice the size of Texas. Conservationists believe there are between 3,900-6,500 snow leopards in the wild.
“It’s an undisputed fact that no one knows what the numbers are,” said Rhetick Sengupta, president of the Snow Leopard Trust’s board of directors and a principal program manager at Microsoft.
Advocates fear that important decisions about the animals’ protection are being made based on too little data. Sophisticated new analytical tools could help.
“You can do predictive modeling as to where to put your conservation dollars,” Sengupta said.
The seed grants from AI for Earth are small — generally around $5,000-15,000 — but there are larger sums being doled out for bigger projects. Some of the snow leopard work was done pro bono before the initiative started, and the group now has a $15,000 grant.
As part of their $10,000 AI for Earth project, scientists studying endangered Puget Sound-area orcas are gathering together data and images from numerous nonprofit, governmental and academic researchers and putting them into the cloud to facilitate collaboration. Up to this point, the scientists were emailing giant attachments or passing around USB drives.
Each of the different research groups collect unique information on the orcas, including aerial photos indicating their weight, blubber samples that measure toxic contamination, fecal samples tested for pregnancies and surveys of marine pollutants. Pulling all of these pieces together could create detailed health profiles of the orcas, which currently number 76 whales.
Conservationists are trying to save the orcas, whose population has been affected by declining numbers of Chinook salmon, their preferred prey, as well as exposure to pollutants and vessel traffic and noise. Because there are numerous factors in play, it can be difficult to know what effect reducing salmon fishing in a certain location will have, or if restricting boat traffic in the orcas’ feeding area could help.
“Those are simple questions to ask when the data is centralized and you have the computing ability,” said Joe Gaydos, science director for the nonprofit SeaDoc Society.
Gaydos is coordinating the data compilation effort. And while it’s too soon to know the impact of putting more technical muscle behind conservation efforts, Gaydos expects a sea change for the field.
“It’s going to be a paradigm shift for the management of wild populations,” Gaydos said.
But many of the environmental organizations are cash-poor and lack tech expertise. Joppa says his team is mindful of these limitations, and that the technology shouldn’t be cost or skill prohibitive.
“The great thing about machine learning is that a lot of the costs are upfront,” he said. “It’s about getting the data together and getting the model trained up.”
Updates and changes will be less expensive, and Microsoft plans to keep developing and fine tuning tools for a variety of organizations to use. The power of the cloud is that it can be used strategically to keep costs down.
Joppa, who previously worked as a computational ecologist for Microsoft Research for more than seven years, knew some of the players in the enviro field before launching AI for Earth.
“We know who a lot of those people are, and we’ve been waiting for the technology to be good enough to be able to actually help them,” he said. “And we feel like that’s where we are right now.
“It wasn’t like this topic sat in Research for 10 years because nobody cared. It sat there because it wasn’t ready and now it’s ready,” he said. “It’s just ready — but it’s ready.”