Recently, with the powerful assistance of Anthropic's large model Claude, astrophysicists have successfully produced the first complete ultraviolet all-sky map in human history. In this vast star map containing 119 million stars, roughly one third of the sky was inferred and completed by Claude based on past multi-band astronomical observation data, an achievement that successfully filled a historical gap left over the past 50 years because ultraviolet survey telescopes deliberately avoided the bright regions of the Milky Way.
In astronomical research, producing an all-sky map is an extremely arduous and complex systems engineering effort. The grunt work that originally required astrophysicists to spend weeks on manual processing has seen a leap in efficiency with the intervention of AI. By flexibly orchestrating multiple agents, Claude efficiently completed the entire workflow of data collection, calibration, cross-calibration, and inference-based completion, dramatically shortening what had been a weeks-long cycle to just a few days.
Even more astonishing, after rigorous comparison, the error between the AI-completed regions' data and actual measured values was maintained at only around 10%. To ensure scientific rigor, the model also precisely annotated in the final image, for every pixel, whether it was measured or inferred, along with the corresponding error range.
This groundbreaking practice shows that fundamental research problems long backlogged due to limited manpower are gradually being handed over to AI agents. In future frontier exploration, human researchers can extricate themselves from tedious low-level data processing and focus more of their energy on steering research direction and identifying core scientific questions.