Artificial Intelligence and the Classification of Prehistoric

Making sense of variation in ancient decorated pottery plays a critical role in Southwestern archaeology, both in chronology (dating archaeological sites) and cultural affiliation (assessing and tracking the identities of ancient populations across space and through time). In this talk Leszek Pawlowicz and Chris Downum present the historical background of pottery typology in the American Southwest, the use of pottery types for archaeological dating, and some of the problems that archaeologists have when assigning types to individual ceramic sherds. They ask a simple question: Are machines (computers that can analyze digital images of pottery) potentially more reliable than humans when it comes to identifying ceramic types? They will explore the answer by presenting results of recent experiments evaluating the utility of deep learning / artificial intelligence to the classification of ancient decorated pottery fragments from Northern Arizona.