The Herculaneum scrolls, carbonized by the eruption of Mount Vesuvius in 79 CE, are being deciphered using cutting-edge technology and artificial intelligence. Scientists are employing a combination of X-ray imaging, machine learning, and computer vision to virtually unwrap and read these ancient texts without physically touching them12. The process involves:

• Scanning the scrolls using synchrotron X-ray beams to create 3D reconstructions1

• Applying AI algorithms to detect ink and enhance text clarity23

• Using a segmentation approach to improve coherence and visibility of text lines4

• Collaborative efforts through the Vesuvius Challenge, which has awarded over $1 million in prizes to participants5

This groundbreaking work has already yielded results, with the first word "porphyras" (meaning "purple" in ancient Greek) successfully deciphered by a 21-year-old student6. The ongoing research promises to unveil a wealth of previously unknown ancient knowledge, potentially expanding our understanding of Roman philosophy and literature67.

TimeSformer, a video architecture based purely on Transformers, played a crucial role in deciphering the ancient Herculaneum scrolls. Developed by Facebook AI, TimeSformer excels at capturing long-range dependencies in video data, making it ideal for analyzing the complex patterns in the x-ray scans of the scrolls12. The winning team of the Vesuvius Challenge utilized TimeSformer as part of their ensemble of AI models to detect ink in the x-ray images, enabling them to read approximately 5% of the letters in one scroll3. This breakthrough demonstrates the potential of advanced AI models in unlocking historical mysteries and accelerating research in fields such as archaeology and ancient literature studies24.

AI is revolutionizing archaeological research through various innovative techniques. Machine learning algorithms are being used for predictive modeling, helping archaeologists identify potential excavation sites by analyzing patterns in historical, geographical, and environmental data1. This approach significantly improves efficiency and cost-effectiveness in site discovery. Additionally, AI-powered image recognition technology is transforming artifact analysis, enabling rapid classification and dating of objects with high accuracy2. These advancements have led to groundbreaking discoveries, such as the recent deciphering of ancient Herculaneum scrolls carbonized by the Vesuvius eruption, where AI helped distinguish ink from papyrus and decipher faint Greek lettering34.

• LiDAR data analysis: AI algorithms process LiDAR scans to reveal hidden archaeological features beneath dense vegetation5.

• Drone-based aerial surveys: AI enhances photogrammetry, creating detailed 3D models of sites from drone imagery5.

• Pattern recognition: AI systems analyze vast datasets to uncover connections and patterns in archaeological findings2.

• Non-invasive exploration: AI-driven techniques allow for site investigation without physical excavation, preserving delicate artifacts and structures1.

Luke Farritor, a 21-year-old computer science student at the University of Nebraska-Lincoln and SpaceX intern, made headlines by becoming one of the co-winners of the $700,000 Vesuvius Challenge Grand Prize12. Farritor's breakthrough came when he successfully used artificial intelligence to decipher the first word, "porphyras" (meaning "purple"), from the carbonized Herculaneum scrolls34. His innovative approach involved:

• Developing a machine learning model to separate carbonized ink from carbonized paper1

• Training the AI on scattered letters found across the scrolls4

• Collaborating with other researchers to decode entire passages of text2

This achievement not only earned Farritor worldwide recognition but also demonstrated the potential of combining young talent with cutting-edge AI technology in unlocking ancient mysteries15.