I’m an ecologist and an explorer by nature. I have been lucky to spend most of my time exploring both physical and digital wilderness. When I was young, my parents studied humpback whales in Bermuda, where they discovered that humpbacks sing, and recorded “Songs of the Humpback Whale”. Later, they studied right whales in Patagonia, where we spent several enchanted years over the course of a decade.

After college, my wife, Ann Edwards and I ran a small development project in the Namibian Kalahari for two years, helping a group of Ju/wa Bushmen people to build a sustainable economy after they lost their hunter-gatherer lifestyle. My mother had discovered that elephants communicate at frequencies below the range of human hearing, and she studied elephants in Namibia after we had left, while Ann and I hitchhiked from Cape Town to Nairobi. Back in the US, we joined a tropical ecology program at the University of Florida, and while doing our Masters research on primates and ungulates, we lived for a year in a very remote camp in the rainforest of southwestern Cameroon. When we finished, we moved to New York and I worked for three years for the Wildlife Conservation Society as the Asia Program Officer. In that job, I had the opportunity to spend a few months in the wilderness of Laos, Papua New Guinea, and far eastern Russia. I also witnessed the fact that conservation battles often came down to disagreements over population estimates, and I eventually resigned and started a quantitative PhD program, to learn population modeling.

Population models have historically been most advanced in the field of fisheries management because it requires sophistication to count animals that you can’t see, so I studied fisheries and worked for Ray Hilborn on stock assessment models during my PhD, and later did a postdoc at NOAA, writing extinction risk models for salmon with Paul McElhany. One of the weak points of population models is the need to account properly for spatial movements in order to distinguish immigration and emigration from population growth and to manage exploitation effectively. Unfortunately, the technology to track animals, particularly in the marine environment, had been lacking. Several exciting new technological developments occurred while I was doing my postdoc. Afterwards, I jumped on an opportunity to be a senior scientist for the Pacific Ocean Tracking Project, part of the 10-year, $100 million Census of Marine Life. We deployed a huge network of acoustic receivers along the US West coast from California to Alaska, stretching in lines from the shore out to the edge of the continental shelf, for public use. I built a large database, tagged fish and squid, deployed and recovered receivers using commercial fishing boats, and worked on convincing other scientists to use the array and to share their data. Among the project’s successes was tracking tiny salmon smolts more than 2500 km, from far inland up the Columbia River down to the ocean and northwards along the continental shelf to the Gulf of Alaska.

When POST’s funding ended, I ported the data to the new Ocean Tracking Network and Ann got a job running a conservation program in Mongolia, the most landlocked country on earth. I joined her as the project’s Science Advisor and for two years, I tracked satellite-collared wild asses (khulan) vast distances across the Gobi desert with my colleague Petra Kaczensky. We discovered that we could locate very cryptic water sources used by khulan on the basis of patterns in their satellite tracks. I also joined a team doing an aerial survey of wildlife and livestock over 150,000 km² of the Mongolian Gobi Desert. The survey generated more than 100,000 photographs, and it took six people two months working full-time to identify and count animals.

Around that time, I became aware of the immense potential of AI for supporting conservation, and saw a near-term opportunity to use AI to identify animals in aerial surveys. A grant from Microsoft’s AI For Earth program led to the project described in this blog, which is an effort to make aerial surveys easier, cheaper, and more accurate by developing a data processing pipeline centered around cloud-based AI models.

The tremendous threats of climate change, pollution, overexploitation, and habitat destruction are all human problems at the root, and many could be solved if human institutions were more effective at producing outcomes that benefit both people and nature. AI is poised to become a very effective tool in the struggle to fight disinformation and to generate support for the changes we need. In the longer term, I am interested in how AI can be used to educate and influence people in support of a more sustainable relationship between humans and the planet that we all depend on.

For more information about my research, see my ResearchGate profile.

This website is powered by fastpages [^1]. [^1]:a blogging platform that natively supports Jupyter notebooks in addition to other formats.