🐙 FathomVerse—a Mobile Game for Exploring the Underwater World
Scientists from the Monterey Bay Aquarium Research Institute in California have created a MiniROV, an AI-powered underwater robot. Using AI algorithms, it autonomously locates and tracks marine organisms and collects their 3D images.
Everyone can train the MiniROV AI model so that the robot learns to work completely autonomously. Researchers have created a free game, FathomVerse (available for iOS and Android), to speed up the process.
The interface is simple and meditative: the player wanders in the ocean, spots unfamiliar sea creatures, and tries to identify which species they belong to. When multiple players reach a consensus, researchers verify their findings and compare them to the AI's classifications. Users have already identified almost 48,000 marine organisms—about 14% of the images in the database.
In essence, the players help annotate the data to train the AI. Automating this process is nearly impossible—teaching a model to identify an unfamiliar sea animal is akin to classifying an alien species.
Chris Jackett, a specialist in identifying corals and other marine species using AI, notes that the human ability to recognize patterns remains unmatched, and having multiple observers examine the same imagery helps build robust training datasets. Project lead Kakani Katija adds that FathomVerse players may be the first ever to see some previously unidentified marine creatures.
More on the topic:
🐦 Listening to the Sky: How AI Helps Study Birds
🐈 Lost in Translation: Will AI soon help us talk to animals?
