Scientists have developed an innovative app powered by artificial intelligence to identify dinosaurs based on their footprints, which are millions of years old. Professor Steve Brusatte from the University of Edinburgh explains the challenge of matching footprints to their dinosaur creators. “It’s not straightforward. The shape of a footprint depends on various factors, like the type of ground and how the dinosaur walked,” he says.
In a recent study published in the Proceedings of the National Academy of Sciences, Brusatte and his team revealed that previous AI systems relied on labeled footprints. If those labels were wrong, the AI would also be inaccurate. Dr. Gregor Hartmann, the study’s lead author, emphasized that finding a footprint alongside its dinosaur is impossible, which means some labels may be incorrect.
To improve the accuracy, Brusatte and Hartmann’s team used a new approach. They trained their AI on 2,000 unlabeled footprint silhouettes. The AI identified key features, like toe spread and heel position, to determine how footprints differ from each other.
Their findings led to the creation of the free app called DinoTracker. Users can upload a footprint silhouette and see the seven most similar footprints. The app also allows users to adjust different features to explore how they change the similarity.
Hartmann noted that the AI’s classifications aligned with expert opinions around 90% of the time, though experts still need to verify factors like the material and age of the footprints.
Interestingly, their research provided insights suggesting that some footprints from the Triassic and early Jurassic periods appeared strikingly bird-like, despite being much older than the first known bird skeletons. Brusatte pointed out that if these tracks were made by birds, it would push back their ancestry by tens of millions of years. However, he suspects these footprints might belong to meat-eating dinosaurs with bird-like feet rather than true birds.
Dr. Jens Lallensack from Humboldt University of Berlin, who has also utilized AI in this field, raised a caution about relying too heavily on the new system. He explained that some critical features identified by the AI are not directly tied to the foot shape itself. Lallensack added that bird-like tracks might stem from how theropad feet interacted with soft ground rather than indicating the early emergence of birds.
Overall, the combination of AI and paleontology is paving new paths in understanding dinosaur footprints and their implications for the history of bird evolution. The ongoing debate highlights the complexity of tracing ancestry through fossil evidence and the exciting potential of technology in uncovering these ancient mysteries. For more authoritative insights, check out the related report on the latest discoveries in paleontology.
