NVIDIA GPU and AI Used To Decipher Ancient Roman Scrolls

As many of us are aware, graphics cards have expanded in usecases with the growing AI market which NVIDIA is dominating with their H100 GPUs. Most of the usecases are for complex and mostly boring things but something that is a bit interesting is a recent competition to extract readable content from ancient roman scrolls which has been achieved using a GTX 1070.

Herculaneum Scrolls Deciphered By GTX 1070

As shared on NVIDIA’s Blog, Luke Farritor, a 21 year old computer science undergrad at the University of Nebraska-Lincoln alongside some others, has managed to use a GTX 1070 to decipher ancient texts. The texts are the Herculaneum scrolls which is a library of ancient texts that were carbonized and preserved by the eruption of Mount Vesuvius in 79 AD. Farritor made use of the GTX 1070 and the ResNet deep learning framework and became the first to decipher 10 letters from a small patch of scroll which earned him a $40,000 prize. The text deciphered was “πορφυρας,” which translated to “Purple Dye” or “Cloths of Purple”.

Various other people are involved in this including Nat Friedman, CEO of GitHub and organizer of the Vesuvius Challenge and W. Brent Seales, Chair of the University of Kentucky Computer Science Department who has spent over a decade developing methods to digitally unfurl and read the delicate scrolls. The stakes are also much higher as private donors are offering up to a $700,000 prize for those who can retrieve four distinct passages of at lease 140 characters this year.

If more of this can be achieved with AI, then i’m all for it as there is so much still to learn from human history.

If you would like to learn more about the Vesuvius Challenge, visit scrollprize.org.