Ted Hisokawa
Mar 04, 2025 08:58
A new AI model, Atlantes, developed by Ai2, enhances monitoring of global fisheries and wildlife, aiding conservationists in protecting natural resources from illegal activities.
Innovative AI Model Targets Illicit Fishing
In a groundbreaking development for conservation efforts, researchers have introduced an open-source AI model named Atlantes, designed to monitor global seafaring vessels. Developed by the Allen Institute for AI (Ai2) in Seattle, this model aims to combat illegal fishing by analyzing over five billion GPS signals daily from approximately 600,000 ocean-going vessels, according to NVIDIA.
High Accuracy and Real-Time Alerts
Atlantes boasts an impressive prediction accuracy of around 80% in determining a vessel’s activity. Integrated into Ai2’s Skylight maritime monitoring platform, it can alert authorities within 15 minutes of detecting potential illegal fishing activities. This capability was demonstrated when Argentina’s Navy intercepted and fined a vessel for illegal fishing, following an alert from Skylight.
Technological Backbone of Atlantes
The AI model, consisting of 4.7 million parameters, is built on NVIDIA H100 Tensor Core GPUs and PyTorch. It processes Automatic Identification System (AIS) data, which is mandatory for most vessels, from January 2022 to June 2024. The training involved maritime experts annotating over 15 million signals to enhance the model’s precision.
Wider Implications for Global Fisheries
Illegal, unreported, and unregulated (IUU) fishing costs the global economy up to $23 billion annually, accounting for about 20% of the world’s fisheries catch, according to the Financial Transparency Coalition. Particularly affected are African waters, where local communities heavily depend on fishing for sustenance and employment.
Extending AI to Wildlife Conservation
Ai2 plans to expand the use of Atlantes beyond maritime applications, integrating it with EarthRanger, a platform that aggregates data from various sources to monitor wildlife. This includes tracking elephants, rhinos, and wild dogs to mitigate human-wildlife conflicts. The system will be trained to predict elephant behavior, helping reduce clashes between elephants and farmers.
Future Prospects and Conservation Impact
By utilizing extensive datasets of elephant movements, Ai2 aims to address human-wildlife conflicts proactively. Jes Lefcourt, director of EarthRanger, emphasized the potential to save elephant lives by predicting their movements and preventing conflicts with humans.
The infrastructure supporting the classification of fishing vessels is also applicable to predicting elephant behavior, showcasing the versatility of AI in conservation. This innovative approach could significantly enhance the protection of natural resources and biodiversity worldwide.
Image source: Shutterstock
