Unraveling the Past with AI by Dr. Timothy Smith

Photo Source: Unsplash

Some mysteries of the past have remained unsolvable until a convergence of technologies emerges to give historians, researchers, and anthropologists the tools to glimpse the previously unseeable. For example, the advances in DNA sequencing and archeological recovery of ancient DNA, or aDNA, have revealed that contrary to earlier assumptions, early humans did not just coexist with other human species but interbred with Neanderthals and Denisovans. Based on DNA studies, we now know that European and Asian humans carry about 2% of Neanderthal DNA, and Pacific and Southeast Asians have up to 6% Denisovan DNA. (tandfonline.com) Such findings tell a story that could not have been deduced from archeological digs alone.

The recent developments in artificial intelligence combined with other technologies have produced startling results that have changed how we see the past because AI can identify patterns that the human eye simply cannot see. A remarkable feat of research and engineering has mathematically unrolled and mapped charred scrolls from an ancient library. These scrolls look like charcoal briquettes and come from the town of Herculaneum near its more famous neighboring town of Pompeii in Italy. When Mt. Vesuvius erupted in 79 AD, it swept the region in searingly hot gas and covered the surrounding country and towns in volcanic ash. The eruption occurred so quickly that many people and animals were killed instantly, and their corpses were forever locked in their poses and cast in ash, only to be unearthed thousands of years later by archeologists. Archeologists working in the town of Herculaneum in 1752 uncovered a library called Villa of the Papyri. This library represents the only intact library from the Greco-Roman period in the world. However, its over 1,800 scrolls had been charred by hot gases from the volcanic eruption. Over the past 250 years, researchers have tried to delicately unroll some of the scrolls, but their efforts mostly yielded a pile of ash.

A combination of technologies has recently proven successful in analyzing the scrolls with a technique called High Resolution X-Ray Imaging and an AI technique called deep learning. High Resolution X-Ray Imaging creates a detailed 3-dimensional image of the carbonized scroll down to the resolution of the paper fibers. Imagine if you had a clear sheet of plastic with text written on it and rolled up. In the rolled-up form, a 3D map of the scroll would still not easily yield readable text. Now imagine the text rolled up and both the paper and the ink have appeared burnt and indistinguishable. The first step after building a 3D image of one of the Herculaneum scrolls was to use computer transformation to virtually unroll the scroll to analyze for text. Since both the paper and ink were carbon based very little of the black ink could be deciphered from map of the blackened paper. However, subtle changes in the papyrus made by pressing down by the original scribe created patterns that a human eye could not see, but a computer using deep learning recognized these subtle “crackle” patterns, which revealed the writing hidden in the scroll. Deep learning will dismantle signals from paper fibers and crackle patterns to pull subtle images from what looks like a charred piece of paper. (neh.gov)

The first analyzed scroll is cataloged as PHerc. 1667 “discusses the nature of human knowledge, the ability to distinguish good from evil, and the merits of acting on reason rather than on impulse—ideas central to the philosophy of Stoicism.” (smitsonianmag.com) The demonstration of the high-tech mapping of the charred remains of the scroll from Herculaneum illustrates the potential to look into fragile even inaccessible documents in a non-invasive way that opens the door to a rich library seemingly lost forever. The devastating blast from Vesuvius almost 2,000 years ago preserved the only Greco Roman library existing today, and technology has opened up the text seemingly sealed forever.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.