The discovery at a site in Western Israel is only the sixth location worldwide of evidence of fire over half a million years old.

AI has helped researchers from the University of Toronto, the Weizmann Institute of Science and Hebrew University establish that ancient humans, or hominins, used fire at least 800,000 years ago at an archaeological site in Western Israel. (University of Toronto)

Evron Quarry archaeological site in Western Galilee

The researchers used AI to analyze the artifacts unearthed at the Evron Quarry archaeological site in Western Galilee in the mid-1970s. These included animal fossils and Paleolithic tools dating to between 800,000 and one million years ago.

Visually these artifacts showed no evidence of their use with fire. However, the researchers used AI and spectroscopy in their development of advanced computational models to investigate their chemical and molecular structures, and whether they had been altered by fire or heat.

When they applied these techniques to 26 flint tools found at the site, the team found that they had been heated to a wide range of temperatures higher than 600 C. An elephant tusk also exhibited similar evidence of heating.

“This evidence of fire is from a period very early in human development and suggests that by one million years ago human engagement with fire might have been more common than we thought,” said Michael Chazan, a professor in the department of anthropology in U of T’s Faculty of Arts & Science and a co-author of the study. “What’s more, this method allows us to detect traces of fire where they could not be seen before.”

“The advantage of AI is that it can find hidden patterns across a multitude of scales,” says Azuri, who led the development of the models used on testing the Evron Quarry artifacts. “By pinpointing the chemical composition of materials down to the molecular level, the output of the model can estimate the temperature to which the stone tools were heated, ultimately informing us about past human behaviors.”

Related Story: Now AI Could Unlock Secrets From 2,500-Year Old Iranian Tablets