AI Cracks Dinosaur Footprint Mystery

An advanced artificial intelligence system has achieved approximately 90% accuracy in classifying dinosaur species based on fossilized footprints, a feat comparable to human expert identification. Researchers from the University of Edinburgh and Helmholtz-Zentrum developed this AI by feeding it 2,000 unlabelled footprint silhouettes. The system analyzes eight distinct features to determine similarities, overcoming challenges posed by varying substrate conditions that can alter a footprint's appearance.

The new technology has been integrated into a free application called DinoTracker. This app enables users to upload images of dinosaur footprints, compare them with the seven most similar imprints in the database, and understand how subtle variations in features impact classification. The AI's findings support existing paleontological observations, such as remarkably bird-like footprints from the Triassic and early Jurassic periods, predating the oldest known bird fossils by tens of millions of years.

This AI-driven classification offers a more objective approach to identifying dinosaur tracks, as traditional methods sometimes relied on potentially flawed prior identifications. The system's accuracy provides stronger evidence for evolutionary hypotheses, such as a deeper ancestry for birds or the existence of meat-eating dinosaurs with very bird-like feet. While some experts note limitations, the AI represents a significant advancement in paleontological research tools.