Burnt to a crisp, the multitude of fragile ancient scrolls would disintegrate at the slightest touch, rendering any script nearly indecipherable. Despite their unopened state, the Herculaneum scrolls have yielded their secrets, thanks to the remarkable capabilities of artificial intelligence. In 2023, a team of three researchers utilized AI and high-resolution X-rays to decipher over 2,000 characters from these tightly rolled documents—marking the first time full passages from papyri that survived the eruption of Mount Vesuvius in AD 79 have been revealed.

These artifacts, believed to be from a building associated with Julius Caesar's father-in-law, offer an unparalleled trove of knowledge about ancient Rome and Greece. The Vesuvius Challenge, a competition aimed at hastening the deciphering process, has set a goal for computer scientists to unlock 90% of four scrolls by the close of 2024. The primary obstacle has been to virtually flatten the documents and differentiate the black ink from the charred papyri, thereby making the Greek and Latin script legible.

"AI is enhancing our ability to discern the ink's evidence," explained Brent Seales, a computer science professor at the University of Kentucky, who has been dedicated to decoding the scrolls for over a decade. "The ink's evidence is present, hidden and disguised within the complexity that AI simplifies and condenses." This project exemplifies the expanding utility of artificial intelligence, which gained recognition in 2024 when the Nobel committee acknowledged AI's development and application in science for the first time: The physics prize honored John Hopfield and Geoffrey Hinton for their foundational discoveries in machine learning, which laid the groundwork for contemporary AI usage.

AI, a term often clouded in ambiguity and overstatement, aspires to emulate human cognitive functions to address problems and accomplish tasks. It encompasses a variety of computational techniques, including employing data sets to train and refine machine learning algorithms, enabling them to identify patterns and inform predictions. While some AI tools pose risks, such as those used in hiring, policing, and loan applications that can perpetuate bias by being trained on historical data reflecting prejudiced notions, AI has revolutionized the scientific discovery landscape, with a sharp increase in peer-reviewed papers employing AI tools since 2015. These papers are more likely to be among the most cited. Over half of the 1,600 scientists surveyed by Nature anticipated AI tools to be "very important" or "essential" to research practice.

However, the UK's Royal Society has cautioned that the black-box nature of many AI tools hinders the reproducibility of AI-based research. For Seales, it's the judicious application of a powerful tool that has yielded remarkable outcomes. "AI is a field within computer science designed to solve problems in ways previously thought to be the exclusive domain of humans," Seales stated. "I view the AI we're employing as a superpower that allows you to perceive elements within data that would otherwise be invisible to the human eye." The Vesuvius Challenge is just one instance of how the rapidly evolving field has disrupted science and unveiled the unexpected in 2024. AI is also enhancing scientists' comprehension of how animals communicate in the ocean's depths, assisting archaeologists in discovering new sites in remote and inhospitable terrains, and tackling some of biology's most formidable challenges.

Deciphering the Languages of the Ocean's Depths

Researchers are aware that the enigmatic clicks produced by sperm whales vary in tempo, rhythm, and length, but the meaning behind these sounds—generated by the spermaceti organ in their bulbous heads—remains enigmatic to human ears. Machine learning, however, has facilitated the analysis of nearly 9,000 recorded click sequences, known as codas, representing the voices of approximately 60 sperm whales in the Caribbean Sea. This work may one day enable humans to communicate with these marine creatures.

The scientists scrutinized the timing and frequency of codas in solitary whale utterances, in choruses, and in call-and-response exchanges between the marine behemoths. When visualized with artificial intelligence, previously unseen coda patterns emerged, which the researchers likened to phonetics in human communication. In total, the program identified 18 types of rhythm, five types of tempo, three types of rubato, and two types of ornamentation—an "extra click" added at the end of a coda within a group of shorter codas.

These features could all be combined to form an "enormous repertoire" of phrases, as reported by the scientists in May. However, the approach has its limitations. While machine learning excels at recognizing patterns, it does not elucidate their meaning. A subsequent step, according to the study, involves interactive experimentation with whales, along with observations of whale behavior, which could be crucial in deciphering the syntax of sperm whale click sequences. This approach could also be applied to vocalizations by other animals, as Dr. Brenda McCowan, a professor at the University of California Davis School of Veterinary Medicine, previously informed. She was not involved in the study.

AI-Powered Archaeological Discoveries

On land, artificial intelligence is now accelerating the search for enigmatic lines and symbols etched into the dusty ground of Peru's Nazca Desert, which archaeologists have spent nearly a century uncovering and documenting. Often only visible from above, the vast pictograms depict geometric designs, humanlike figures, and even an orca wielding a knife.

A team of researchers led by Masato Sakai, a professor of archaeology at Japan's Yamagata University, has trained an object detection AI model with high-resolution imagery of the 430 Nazca symbols mapped as of 2020. The team included researchers from IBM's Thomas J. Watson Research Center in Yorktown Heights, New York. Between September 2022 and February 2023, the team tested the accuracy of its model in the Nazca Desert, surveying promising locations by foot and with the use of drones. The researchers ultimately "ground truthed" 303 figurative geoglyphs, nearly doubling the known number of geoglyphs in a matter of months.

The model was far from perfect, suggesting an astonishing 47,000 potential sites from the desert region, which covers 629 square kilometers. A team of archaeologists screened and ranked those suggestions, identifying 1,309 candidate sites with "high potential." For every 36 suggestions made by the AI model, the researchers identified "one promising candidate," according to the study. Nevertheless, AI has the potential to make significant contributions to archaeology, particularly in remote and harsh terrains such as deserts, even though the models are not yet entirely accurate, said Amina Jambajantsan, a researcher and data scientist at the Max Planck Institute of Geoanthropology's department of archaeology in Jena, Germany.

Jambajantsan wasn't involved in the Nazca research but uses an AI model to identify burial mounds in Mongolia based on satellite imagery. "The problem is archaeologists don't know how to build a machine learning model and data scientists, typically, are not really interested in archaeology because they can get much more money elsewhere," Jambajantsan added.

Unraveling the Building Blocks of Life

AI models are also assisting researchers in understanding life at its most minute scale: strings of molecules that form proteins, the building blocks of life. While proteins are constructed from only around 20 amino acids, these can be combined in nearly infinite ways, folding themselves into highly complex patterns in three-dimensional space.

These substances help form hair, skin, and tissue cells; they read, copy, and repair DNA; and they aid in carrying oxygen in the blood. For decades, decoding these 3D structures has been a challenging and time-consuming endeavor involving the use of delicate lab experiments and a technique known as X-ray crystallography. However, in 2018, a revolutionary AI-based tool emerged.

The latest iteration of the AlphaFold Protein Structure Database, developed by Demis Hassabis and John Jumper at Google DeepMind in London, predicts the structure of almost all 200 million known proteins from amino acid sequences. Trained on all known amino acid sequences and experimentally determined protein structures, the database functions as a "Google search," providing access at the touch of a button to predicted models of proteins, accelerating progress in fundamental biology and other related fields, including medicine.

The tool has been utilized by at least 2 million researchers worldwide. "It’s really a stand-alone breakthrough solving a traditional holy grail in physical chemistry," Anna Wedell, a professor of medical genetics at Karolinska Institutet in Sweden and a member of the Royal Swedish Academy of Sciences, stated after Hassabis and Jumper were among the three winners of the 2024 Nobel Prize for chemistry.

The tool does have some limitations. Attempts to apply AlphaFold to proteins based on mutated sequences, including one linked to early breast cancer, have confirmed that the software is not equipped to predict the consequences of new mutations in proteins. AlphaFold is only the most prominent of a number of AI tools being deployed in biomedical fields. Machine learning is accelerating efforts to compile an atlas of every single type of cell in the human body and discovering molecules that become new drugs, including a type of antibiotic that may work against a particularly menacing drug-resistant bacteria.