Counting elephants...

If you were born and brought up in Kerala, any time in the last century, unless you were living in a hole, the name Guruvayur Keshavan is probably familiar to you. Like so many children born before and after his death, I grew up on a steady diet of Gajarajan Keshavan (Gajarajan literally means elephant king) stories that were a mélange of legend, lore, and fairytale. Glorified to an almost superlative degree, it was very difficult to not be fascinated by the aura surrounding the larger-than-life tusker from the famous Guruvayur temple, known for his exceptional intelligence and compassionate nature. As I grew older, the obsession with the jumbos from my picture books simply grew with me, along with my collection of books on elephants. 

Many, many years later, when I traveled to South Africa, I spent painstaking hours explaining to anybody who would listen why Indian elephants were better looking than their African cousins. Very few people agreed with me. But then a lot of them had never seen pictures of the magnificent Keshavan.

This picture doesn’t do justice to his majestic good looks; though I’ve never seen him, I just know.

On a more serious note, what really struck me about a lot of people I met in South Africa was the same near-reverence which I had gotten used to in Kerala, every time we discussed elephants. People described them as strong, magnificent and wise – it was almost spiritual. There was one particular story I heard from a ranger, about a 60-year old matriarch named Amahle, who unerringly took her herd to new, forgotten watering holes every drought season. The ranger said it was like she had inbuilt GPS, with maps from everywhere she had been, in her entire life. What that means is, in times of scarcity, most matriarchs would know exactly what to do to help their herds survive. And not only their herds, the entire wildlife ecosystem around them would benefit from the matriarch’s knowledge. For more reasons than just this one, elephants are a keystone species.

That’s probably why, in the natural order of things, elephants who make it to adolescence, have an average life expectancy of about 70 years, unless they’re unlucky enough to be hunted by lions. So the biggest threat elephants today face is a species that hunts and kills for no justifiable purpose.

And no matter how many trackers, fences and rules, conservationists and governments set up, the poachers simply go wherever the elephants are. They poach GPS-collared elephants in Myanmar – not just for ivory, but also for meat and skin, thus putting even the tuskless females and calves at risk. They hound the critically endangered elephants in Sumatra.

They kill African savanna elephants, the world’s largest terrestrial mammal, in Bostwana, Africa’s last elephant safehold. Even the world’s smallest elephants, the Borneo pygmies, are not safe from people.

At the turn of the 20th century, there were a few million African elephants and about 100,000 Asian elephants. Today, there are an estimated 450,000 - 700,000 African elephants and between 35,000 - 40,000 wild Asian elephants. And these precariously low numbers keep falling, because poachers today have become high-tech, using night vision goggles and GPS to track animals and hunt them down.

Rescued baby elephants at the Kottoor Kappukadu Elephant Rehabilitation Centre, Kerala, India.

That’s why conservationists and technologists are working together, more closely than ever before, because there’s an overwhelming realization that technology could play a significant role to help save the elephants, and other wildlife. Microsoft’s AI for Earth program is one such program geared towards exactly that – it puts the power of Microsoft cloud and AI tools in the caring hands of environmental scientists, activists and others trying to solve burning problems around climate, water, biodiversity and agriculture.

Probably the most prominent built-for-AI use case is image analysis using deep learning techniques. Conservationists across the globe collect images from drones and planes and install camera traps, embedded with motion-trigger sensors, that are designed to take photos when they sense movement. This is an important part of counting animals to save them. But there are millions of images to analyze, and a lot of these images are useless, particularly from camera traps, where the empty images can be as high as 75%, according to data from the world’s largest camera-trap project.

Computer vision can solve this problem very effectively – right from eliminating the empty images to identifying species, counting animals, and even observing animal behavior and attributes. In fact, Seattle’s Snow Leopard Trust has built something along these lines for snow leopard tracking, where computer vision algorithms are being used to identify animals in camera images, with high degrees of accuracy.

Another equally great use case is smart patrolling. Data about the nature and features of patrol routes, incidents of poaching and their geographical distribution, frequency, time of occurrence and more can all be used to build and train anti-poaching patrol models, that can be used to predict where poaching might occur in the future. This is currently being used to create optimized patrol routes for rangers in Malaysia and deployed in Uganda’s Queen Elizabeth National Park, as well as wildlife preserves in South Africa.

PAWS or Protection Assistant for Wildlife Security, one of the most popular AI driven conservation tools for wildlife, in general, and elephants, in particular, goes one step further. It’s a clever combination of machine learning and game theory – it uses data from rangers and patrols, to predict where poaching is likely to happen, and then comes up with randomized patrol routes so poachers don’t know where rangers are going to be. With successful implementations in Uganda, Malaysia, China and more, it’s a powerful tool in the war against people who kill animals to fund terrorism and civil wars.

But personally, what I’m really stoked about is the Elephant Listening Project, based at Cornell University’s Lab of Ornithology, that tracks forest elephant herds throughout the Congolese Nouabalé-Ndoki National Park, using about 50 sensors distributed across the park. The good folks at Cornell call it “eavesdropping” on elephants. I just love the idea! The idea is to distinguish between forest elephant calls and all the other sounds in a noisy tropical rainforest. The sensors generate seven terabytes of data every three months. For a human being, it’s like listening to a few million songs without understanding the lyrics. It’s beautiful, but pretty difficult to make sense of.

For AI, it’s perfect – looking for rare patterns in terabytes of data that would take humans years. This elephant call data analysis is used to estimate herd populations more accurately, track herd movements, and even individual elephants who can’t be seen from the air, because of the dense forest cover, and identify poacher chatter in the background.

The voice data analysis, which used to take six to eight weeks to generate meaningful insights, is now possible is three weeks, thanks to algorithms from Conservation Metrics, built with the help of a Microsoft AI for Earth grant. And with the power of the Azure platform and Microsoft’s resource support, this timeline could drop to a single day, very soon. Which means, rangers will then have a real shot at actually stopping an elephant from being captured and killed.

This is just the beginning of AI-driven possibilities and use cases that probably haven’t even been conceptualized before. In the Serengeti, conservationists are piloting AI-equipped cameras that can work as remote lookouts, tools like AirSim are being used by researchers to simulate animal-habitat images to quickly build custom datasets to train computer vision models faster, crowdsourcing in the form of pictures taken by citizen scientists is contributing to cataloguing local wildlife which computer vision models then help label..

Every 15 minutes, an African elephant is killed. And the Asian elephants are not too better off. If poaching continues at the same rate, the entire remaining population will be wiped out in a few short decades. And elephants like Keshavan and Amahle will be confined to sad bedtime stories about Earth’s lost wealth. 

Can you imagine children growing up in a world where elephants exist only in books? Human beings are all that stand between such a dark world and those that so callously hunt these majestic creatures.

Weapons forged in AI may very well win the war for elephants.

Whoever it was who got to see this adorable baby elephant in real life, long enough to stop and take a snap, I envy you.