The world of artificial intelligence (AI) is evolving, and the most powerful stories are not about chatbots. Instead, they revolve around the quiet algorithms that are revolutionizing various fields, from archaeology to astronomy. These algorithms are not just about fluency; they are about scale and the ability to process vast amounts of data that would be impossible for humans to handle alone. The AI that captivates the public's attention, like chatbots, is often overshadowed by the quieter, more specialized systems that are making groundbreaking discoveries.
One such example is the Vesuvius Challenge, where high-resolution X-ray scans and machine-learning models are being used to read ancient Roman scrolls that were once too brittle to unroll. The project, launched in 2023, has already achieved remarkable results, including the recovery of over 2,000 characters from a scroll, an Epicurean text by the philosopher Philodemus. The model flags the likely locations of ink, and papyrologists then confirm the letters and their meaning. This is a powerful demonstration of how AI can make the unreadable legible, but it is still a collaborative effort.
Similarly, in astronomy, deep-learning models are being used to rank and identify strong gravitational lenses, which are valuable for studying dark matter and cosmology. The Euclid mission, for instance, has already produced a catalogue of 497 galaxy-galaxy strong lens candidates from just six weeks of searching. This is a significant achievement, but it is the human experts who then vet the top-ranked objects, ensuring the accuracy of the results. The pattern is the same: AI sorts and ranks, and humans confirm.
The scientific establishment has already recognized the value of these narrow AI systems, with the 2024 Nobel Prize in Chemistry going to David Baker, Demis Hassabis, and John Jumper for their work on AlphaFold, a system that predicts protein structures. This prize highlights the fact that these AI tools are seen as discovery tools, not just software demonstrations. They are not general-purpose minds, but highly specialized filters that can process vast datasets and identify the most promising candidates for further investigation.
The key to understanding these AI systems is recognizing that they are not about fluency or understanding in the human sense. Instead, they are about scale and the ability to process data that would be impossible for humans to handle alone. They are very good filters, and their value comes from the size of the haystack they can process, not from any understanding of what they find. This distinction is crucial, as it affects how we interpret and use these tools.
The trend to watch is the widening gap between the data being collected and the people available to look at it. As more data is generated, from Herculaneum scrolls to the vast image volumes produced by the Vera C. Rubin Observatory, the assumption is that AI will rank the data first, and humans will then examine the top-ranked objects. This is the future of discovery, and the quiet algorithms are leading the way.
