Claude Science fills gaps in all-sky UV map
An Anthropic researcher used Claude Science's agent teams to merge decades of space-telescope data into a complete ultraviolet map of the sky, with about a third of it predicted by a model rather than observed.
Brice Ménard, an astrophysicist at Johns Hopkins University and a researcher at Anthropic, has described how he used Claude Science to build what Anthropic calls the first complete map of the sky in ultraviolet light. The map combines far-UV (154 nm) and near-UV (232 nm) data.
Earth's ozone layer absorbs UV light, so UV observations have to come from space, and existing coverage has gaps. NASA's GALEX mission (2003–2013) imaged about two-thirds of the sky across roughly 38,000 observations. It deliberately skipped regions with very bright stars, including much of the Milky Way's galactic plane, to avoid damaging its detectors. Data from NASA's Swift and South Korea's FIMS/SPEAR fill in some of those gaps, but not all of them.
According to Ménard, Claude coordinated a team of agents that searched for public UV surveys and downloaded them. The agents corrected glare from bright stars, worked in parallel to make each survey internally consistent, then cross-calibrated the surveys and merged them onto a common resolution and coordinate grid. Ménard says he gave high-level instructions throughout.
To cover the roughly one-third of the sky with no UV data, Claude used inpainting. It also learned how UV brightness relates to visible, infrared and radio observations in the regions that have been mapped, then applied that relationship to the unobserved regions. Each pixel in the map is labelled "measured" or "predicted" and carries an uncertainty estimate.
Ménard says he checked accuracy by testing the model on regions that have been observed, but the excerpt provided cuts off before the results. He presents the map mainly as a teaching tool, and as an example of AI taking on low-priority research work that scientists tend to put off.