The full text is 2592 words in total, and the estimated learning time is 7 minutes
Source: unsplash
I have worked in the most beautiful and harsh environments on all continents. But whenever a new technology is deployed, I still get anxious. The first field test is always difficult for inventors: Will it work like it would in a lab? Did you miss anything? Hope no one sees me pressing the power switch...
This time the project is even more stressful. At the time, I was in a protected area deep in Indonesia's wet jungle, using a set of custom 3D scanning devices assembled for $60,000, trying to create a digital copy of one of the world's last remaining Sumatran rhinos.
This is a great opportunity for us to capture a digital copy of the rhino and help us tell the story about it. Even in the worst case scenario — if it still can't protect its environment, it can help provide a complete digital record of the rhino, and the National Geographic crew filmed every step of our action.
Pictured here is the critically endangered Sumatran rhinoceros, and there are fewer than 80 left in the world. | Image source: shutterstock
My work never stopped, and months of high-tech systems protection lab development, exploration, or anti-poaching security applications were interrupted from time to time by tight site deployments.
A graduate of MIT with a degree in Computer Science and Artificial Intelligence, I didn't think at first my career would go down to exploring and protecting the establishment of technologies and algorithms, and working with National Geographic gave me the opportunity to combine my passion for computer science and artificial intelligence with conservation, exploration, and storytelling.
A three-dimensional digital model of the Sumatran rhinoceros. The Sumatran rhinoceros is one of the most endangered land mammals on earth | Image source: Author 3D scan rendering
As an explorer and National Geographic fellow, I'm inextricably linked to incredible adventures. Unlike most jobs, it's really as cool as it sounds.
I've been on the deck of the Titanic at a depth of 12,500 meters aboard a 3-man submarine, and the battery technology I helped develop powers the powered robots exploring the wreck cabin; last year, I took a complete, high-resolution aerial lidar map of the world's tallest glacier in a helicopter over Mount Everest.
I have worked with my fellow protectors in conservation operations to stray cameras to the wings of a two-seater plane and fly over the Democratic Republic of the Congo, where the front lines are against elephant poaching; I have spent hundreds of hours in pitch-black environments on tasks similar to diving in submerged caves to scan Mayan sacrifice victims; and I have stuffed my body into vertical cracks that are only 7 inches wide in some places, where cracks squeeze my body as I land in search of former human remains with surface penetrating radar.
Three-dimensional digitization of ice age bear skulls in flooded caves, | Source: Gran AcuiferoMaya project
Overall, I'm encouraged and excited about the role that technological innovation plays in sharing and protecting the most beautiful wealth in the world, but it also scares me because what we really care about is too fragile.
Recently, I was named Rolex National Geographic Explorer of the Year 2020. Each year, this prestigious award is awarded to a leading figure in the adventure story who is a symbol of individual action, achievement and spirit. Previous winners include filmmaker James Cameron, underwater cinematographer Brian Skyley, Steve Boyes and his Okavango Wilderness Project team, and environmental anthropologist Kenny Broad.
Today, the National Geographic Society's mission to harness the power of science, exploration, education, and storytelling to illuminate and protect the wonders of our world is more important than ever, and it is a great honor to be a part of that mission.
Rolex National Geographic Explorer of the Year Award
While digitizing The Sumatran Rhino, I returned to my previous accommodation in Indonesia and used my laptop for a visual preview. When I move around the screen with 3D pictures of animals, the authenticity is astounding. When I stopped moving the mouse, I could barely tell that the view wasn't really a photo.
With a dual discipline of artificial intelligence and philosophy, it is difficult for me not to think about what is a "real" image. I certainly think that the photos I take with my phone are real images, but apart from a bunch of numbers that represent pixel values, what exactly is that? If a synthetic digital rhino can't be distinguished from a photograph, is it as real as a photograph?
I realized that if I could create 3D models that were real to me, I could use those images to train AI systems. Data is the Achilles heel of AI, and good AI models require a lot of data, but in some of the most impactful applications, such as protection, anti-poaching, security, and medical imaging, it's hard to get good data.
The idea made me take the plunge. I founded Synthetaic, an AI and synthetic data company focused on the highest risk AI use cases. Due to the limited number of sample islands in these use cases, high-quality predictive modeling efforts are hampered. In many areas, there simply isn't enough data to effectively train networks, especially for still and moving images.
Poachers with rifles deployed in the wild. | Source: Synthetaic
There are few photos of illegal poachers, as are photos of the new Toyota pickup truck with the extremist logo. Because of the limited storage of images of rare brain cancers, real-time feedback from neurosurgeons in the operating room is limited. Due to the widespread problems in training data, racial bias is very common in facial recognition.
I started using Synthetaic to answer the two biggest questions in AI for me. What if the boundary situation no longer exists? What if the training data is no longer a constraint?
Since founding Synthetaic, we have expanded the concept. We realized that by combining 3D modeling and ARTIFICIAL intelligence, we could grow data at a faster rate and at a lower price, and that it was better suited to training state-of-the-art AI than any other data synthesis technique.
Currently, we are producing microscopic images of human tumors for decision-making during brain surgery, for detecting chest X-ray images of COVID-19, for protecting projects to synthesize aerial images, and to create data for some unresolved security intelligence needs. In these cases, the data is no different from the real image, it is generated almost instantaneously on the server, and we can use this data to train AI models to surpass the current advanced level.
Synthetic chest X-ray images for COVID-19 detection. | Source: Synthetaic
We need technology as a force multiplier to solve the world's toughest problems. Even simple technologies that we see now, such as pulleys or electricity, have radically improved the capabilities of individual individuals and species. I believe that AI can become such a tool when it reaches its full potential.
AI will help the few rangers protect vast areas from poaching; it will help doctors diagnose medical problems earlier when more treatment options are on the table; and it will help us stay safe in an uncertain world. To achieve such a vision, we need to input high-quality data infinitely to unlock the "impossible" AI.
Leave a message like attention
Together, we share the dry goods of AI learning and development
If reprinted, please leave a message in the background and abide by the reprint specifications