Lately you may have noticed commercials from different tech companies touting their initiatives that use AI to help protect the planet. As AI and machine learning become more advanced, many tech companies see opportunities to capitalize on growing concerns about global warming and animal extinction as ways to improve their image. (Who can’t love a company that saves fluffy polar bear cubs?) But these initiatives also demonstrate the value and capability of this new technology to potential customers. It’s great advertising—and it’s actually great for the planet, too. Win, win.
But have you ever wondered how this technology works? What exactly does Microsoft, or IBM, or Google, or any number of tech giants do with their technology that helps conservationists?
Table of Contents
Counting Animals Is Key To Conservation
A big part of wildlife conservation is tracking the number of individuals within a population. If populations are shrinking or there are fewer young members in a population, that’s a sign the species might be in danger. However, if a population is growing, that’s a sign that existing protections or environmental factors are good for the species.
In the past, this has required a lot of time and money. Scientists would travel to the place the animals live, places that often lack infrastructure or are harsh environments for humans. Then scientists would tag individuals in a process called capture-recapture. This can be stressful for animals since it involves chasing, capturing, and sedating them. It can also be dangerous for both animals and humans, as well as expensive and time-consuming.
Another method of counting animals is through aerial surveys. In this method, scientists fly over a population and photograph it. Later, they use the images to identify individuals in the population and make calculations about the size of the total animal population. But like capture-recapture, this process is time-consuming and expensive. And with both of these methods, scientists return home with tons of images to sort through before they can begin recording data. This can creates months and months of tedious work, meaning scientists won’t know whether a population is declining for quite some time.
Counting Animals With AI
AI makes this process a lot easier by analyzing the photographs scientists gather. By measuring and making calculations, AI can quickly identify subtle differences in the pattern of spots, the ridges of a fin or tail fluke, or ear outlines on an animal to distinguish one individual from another.
But it isn’t just scientists’ photographs that AI is looking at. Wildbook, an organization using technology and data science to combat extinction, is creating tools to gather information about animal species from social media sites. One example is a bot they created to crawl YouTube searching for videos of whale sharks. Most of these videos are uploaded by unsuspecting tourists or local scuba divers simply wanting to share their encounters, but Wildbook’s bot can use images from the videos to identify whale sharks by their spots. The bots pull location data and time stamps from the videos as well and enter this information into Wildbooks’ database Wildbook for Whale Sharks. This information helps scientists gain a greater understanding of the size of whale shark populations, as well as their migratory patterns. The real value of this is that these bots can find animal sightings that researchers never would because of lack of funding or logistical challenges. The bots can do this quickly, too, combing through 30 videos daily. And the more the bots work, the faster they become.
Open Databases And Citizen-Scientists
Organizations like Wildbook are creating open databases that scientists can use to record information about different populations of various species. Like, Wildbook for Whale Sharks, GiraffeSpotter is a great example. A group of scientists measuring a giraffe population in one part of Africa is no longer limited to just their own data. Instead, they can go to the GiraffeSpotter database and see information about other giraffe populations in other parts of Africa, helping them build more accurate population models, identify new locations where the animals congregate, and even assess whether protected areas are working to rebuild giraffe numbers.
Wildbook’s databases aren’t just used by scientists, either. Concerned citizens are encouraged to upload their own pictures or videos for the bots to analyze. These citizen-scientists contribute tons of useful, free information for scientists to work with. Analyzing all of these images would be impossible without AI to speed things up.
Combatting Poachers Through Predictive AI
Even when land is set aside as sanctuaries for at-risk species, that doesn’t always ensure the animals living there will be safe. Global demand for endangered animal products—such as ivory from elephant tusks, rhinoceros horns, or tiger testicles, which are sold as luxury goods or folk medicine—makes poaching a lucrative business. Tracking poachers can be dangerous and frustrating, but employing AI can make this safer and more effective.
Until recently, rangers had to make predictions about what routes poachers were likely to take. This was a hit-or-miss process, which made it expensive and less effective. However, Professor Milind Tambe and the USC Center for Artificial Intelligence in Society (CAIS) have developed a tool called Protection Assistant for Wildlife Security (PAWS) to help rangers combat poachers. PAWS integrates with another tool, Spatial Monitoring and Reporting Tool (SMART), which stores data about routes poachers have historically used. Utilizing AI and machine learning, as well as the data from SMART, the PAWS software identifies patterns of activity, helping rangers use their resources more strategically.
Although these tools have been used limitedly, they show a lot of promise. In Cambodia, rangers using the tools recovered more than 1,000 snares during December 2018 and January 2019, twice as many as they recovered before the introduction of PAWS, as well as vehicles and tools used by poachers. By the end of 2020, this software could be in effect at over 600 wildlife sanctuaries in 55 different countries.
The Future Of AI And Conservation
AI is positively changing the way we monitor animal populations, providing more data more quickly. And this is just the start.
Currently, programmers are working to enable AI to analyze images from camera traps. Camera traps are remote cameras that are triggered either by movement or an infrared sensor. They are convenient for scientists because they can be set up and left for an extended period. When they are retrieved, these traps provide hundreds of thousands of images of any species that passed through the area, providing data on ecosystems as well as individual animal populations. However, sorting through so many images is time-consuming. Right now, AI struggles to work with these images, which are often blurry or taken in low-light conditions, but Microsoft and National Geographic are currently partnering to find ways to use AI to process these images.
Organizations like Wildbook that provide open databases also have goals for their offerings. Right now, many of these organizations are working toward continuous monitoring of species through live data. They believe that live data will enable conservationists and governments to intervene more quickly when populations begin to drop. Another benefit is that live data will enable regular people to see the effects of global warming, deforestation, and habitat loss on animals, motivating them to make changes that will help protect the planet.
When we see news stories on melting glaciers, rainforest wildfires, and worsening air pollution in many cities, it’s easy to feel hopeless. Our planet it rapidly moving to a point of no return. But initiatives like these give us a reason to hope. By leveraging emerging AI and machine learning, we can better preserve the vast number of unique and glorious species that populate our planet, leaving a more diverse and beautiful planet for future generations.