Let me tell you a story about a rhinoceros named Sudan. He lived in Africa and was protected 24/7 by armed guards. Why? He was one of the last three northern white rhinos left in our world. And last year, he died, leaving behind just two members of a species, both female With that, we just lost another great animal from our planet. If that sounds tragic, things are worse now. Every day, two rhinos are poached somewhere in Africa, and by the time I finish this talk, another elephant would have been killed. How did we get here and when did things go so wrong? To understand that, we have to start right at the beginning of our evolutionary past. If the Stone Age brings up images of our cavemen ancestors living in perfect harmony with nature, you may be in for a shock. Sample this: About 50,000 years ago, for the very first time, humans entered Australia, a continent teeming with wildlife and resources. But within a few thousand years of their arrival, every single megafauna in Australia disappeared The same pattern has repeated continent after continent and island after island. Research has proven that human migration has always preceded mass extinction of animals in every geography. In the book Sapiens, author Yuval Harari calls humans as the deadliest ecological serial killers our planet has ever seen. So it's no surprise that we are in the midst of the sixth mass extinction of our planet. And the worst one since dinosaurs roamed the earth millions of years ago. Yes, there have been several desperate attempts to reverse this trend. Recently, the great elephant census was a massive exercise to manually count African elephants by flying small planes all through the continent. But two years and $8 million later, we neither have results with granular data, nor a scalable initiative.. Unfortunately, every such genuine attempt suffers due to a shortage of people, resources or time. Meanwhile, all this imbalance is pushing our planet to the brink of disaster. Our world is in total chaos, and this is entirely driven by human greed However, just as necessity is the mother of invention, chaos is the fuel to innovation. Where do we see the silver lining? Enter artificial intelligence. When most people hear AI, what comes to their mind? Killer bots from the Terminator? Machines that will take away our jobs? Or even suspicious websites that influence elections? (laughter) Surprisingly, artificial intelligence can play a positive role in saving our planet. Paradoxically from our very own hands. In fact, it's already happening. A branch of AI called "deep learning" has been inspired by the way kids learn from flashcards. When I show my four-year-old daughter a picture of an animal, I call it an elephant. Another picture, I call it a rhino, and I repeat a few times over, voilà, she's able to identify the animal in any setting: as a toy, on the TV or in the forest. Deep learning works quite the same way and has a photographic memory, learning purely from examples and without the need for any explicit instructions. It is pretty smart at automatically identifying patterns from pictures, videos, audio clippings. In the recent years, AI and deep learning have been silently powering some of the most revolutionary changes in conservation. Two years ago, my team worked with a conservation group in Washington to identify and monitor salmon species. What drove our interest in this fish? No, this has nothing to do with the fact that my wife loves them smoked. Once upon a time, the rivers of the Pacific Northwest were so full of wild salmon, that the fishermen like to say that you could cross the river on the fishes' backs. But today, close to half of this population has vanished. So this is a tragic instance which has happened in our lifetime. Salmon are what's known as a keystone species, not only as a fish -- a main source of food for hundreds of predators but their rotting carcasses supply nutrients to the trees. When the salmon are gone, the rest of the ecosystem won't last long. The challenge we faced was to uniquely identify 12 species of salmon from underwater camera videos. We showed the AI program 1,000 videos of the fish This was basically a case study to look at how the AI can learn from looking at the fish And then this AI program started identifying the fish in every single instance without exception. Despite our creator bias, initially, this looked too good to be true. And then we really discovered what was happening. The AI was actually studying the background of the video, without actually studying the fish. So you might have heard of weak students who can memorize an entire book, but fail to answer a single question. This AI was in their league, learning by rote, and not gaining any intelligence. So when you fix the learning process, the AI truly learn to identify the fish almost as accurately as trained biologists. And what took the biologists 100 hours was brought down to less than 20 by the AI. Here is the AI model in action. You can see how it's able to instantly recognize the fish species, and it's able to identify it even when just a portion of its body is visible It gets pretty smart. So this was two years ago, AI has gotten even smarter today. "Wild Me" is a Portland-based volunteer group funded by Microsoft AI for Earth. They work with whale sharks, the largest fish on our planet, about the size of a school bus. Whale sharks travel thousands of miles every year. And the unique fingerprint of whale sharks is the pattern of spots that you see behind the gills. Why is this important? In the past century alone, over two thirds of whale sharks have been wiped out for their meat and oil. So Now that the AI can uniquely identify every single whale shark, and use these spots to distinguish any two apart, what does it do with this intelligence? It has been binge watching, not on Netflix, but on YouTube. (laughter) Watching videos uploaded by tourists and volunteers: over 10,000 whale sharks have been uniquely identified and cataloged from around the world, almost like a census. And now, whale sharks have their own social network. No, I just made this up. They are not on Facebook. (laughter) Thankfully. Their social network is called the "Wildbook." Anyone can go online, make friends with them, and for a few dollars, give them a nickname or even adopt one as an online pet. So AI can recognize the species it can identify unique animals within a species. But how does it save them from the poachers? The illegal wildlife trade is annually worth eight to 10 billion dollars. Africa, a continent rich in biodiversity, supports the biggest casualties. Here poachers have been indiscriminately hurting and firing at the animals for decades. So this has been continuing and in a hurry to finish off entire herds of elephants poachers poison the watering holes in forests with cyanide. This kills not just the elephants, but it decimates the entire ecosystem in the forests. The Lindbergh Foundation runs an AI-driven conservation project in South Africa. Here, drones fitted with night-vision cameras are guided by artificial intelligence and act as an eye in the sky watching over the animals. They detect the animals' herds and identify endangered species, like the elephants they have spotted in this picture with the orange boxes. They also spot the suspected poacher movements and then they predict when and where the poachers will intercept the animals. This intelligence is relayed back to the rangers who can reach the spot even before the poachers can. In an area in the forest, which saw over 20 rhinos killed every month earlier after this pilot was implemented, the killings were brought down to zero. Zero deaths, six months in a row. (applause) Similar spectacular results have been seen with pilots to protect elephants, thanks to the accurate identification and real-time intelligence. What works in Africa today can be replicated around the world at scale. These three examples show how AI can give a new lease on life to the biodiversity on our planet by helping spot, count, and predict the movement of endangered species in real time. They make it easy for our good samaritans to protect the animals from danger. Organizations and universities around the world are democratizing this capability and making such AI applications public and freely available. NGOs or researchers anywhere in the world can readily deploy and put them into action You could join this revolution as well in a number of ways. Firstly, go ahead and share on social media. Yes, but not just your selfies, but your vacation pictures of animals or even local sightings of biodiversity around you. Add them to social websites and environmental apps like iNaturalist. This lets the AI identify and catalog animal sightings that could be new, or even repeat sightings. Secondly, you could volunteer time to label animal pictures. This is a key manual activity needed to train the AI before it can truly get smart and not just fake it. You could do this on websites like PenguinWatch.com. And finally, contributions help fuel the AI that are ever-hungry for computing power. Imagine a world where we humans can live in newfound harmony with our biodiversity. Animals will no longer be just statistics in casualty reports in the news. Very soon we'll have a clear sense of the health of every animal species, be able to take timely actions before danger strikes, and even adopt them as our own extended pets in the wild. As Deepak Chopra said, "Every great change is preceded by chaos." AI and deep learning are those change agents which are transforming the chaos around biodiversity into order. Thank you so much. (applause)
