A radio telescope starts finding signals on its own: one plain-language request lets agents steer the dish and process 2.4 terabits per second in flight
Synopsis
Breakthrough Listen and the SETI Institute have brought AI agents into the daily search at the Allen Telescope Array, where a single natural-language request lets agents steer the instrument, calibrate and focus it, correct data errors and apply AI models to the incoming stream. On NVIDIA's Holoscan platform, the DAQIRI library pulls terabits per second of sensor data straight into GPU memory so observations are processed in flight, and the system handles 2.4 terabits per second on RTX PRO 6000 Blackwell GPUs. The pipeline has achieved the first real-time AI detection of a pulsar, in the Crab Nebula, and the team is now evaluating open models including Nemotron for settings where data locality and turnaround matter most.
Interpretation
The radio search for extraterrestrial intelligence now has AI agents taking part in operating the observation, not only in analyzing data afterward. The prior condition was that scientists had to identify potentially interesting signals as observations unfolded, against the enormous volumes of data radio telescopes generate. At the Allen Telescope Array, agents can steer the instrument, enhance and correct incoming data and apply AI models to the incoming stream, all from a single natural-language request.
Data is processed while it is being collected rather than written to disk first. Previously the massive sensor streams had to fill hard drives before analysis, which worked against identifying signals during the observation. DAQIRI, a Holoscan platform library, pulls terabits per second of sensor data straight into GPU memory, and CyberEther, a signal-processing framework developed at the SETI Institute, does the math with CUDA-X libraries including cuFFT and cuBLAS.
Agents get instrument state, compute resources and pipelines from one orchestrator that acts as their harness. That lets researchers monitor packet loss, read live spectrograms and rewrite signal-processing blocks by prompting in natural language. Nexus, a web-based observatory orchestrator, provides that single source of truth, and the system handles 2.4 terabits per second on NVIDIA RTX PRO 6000 Blackwell GPUs with inference running on premises next to the instrument.
The pipeline has already carried out a real observation task. No real-time AI detection of a pulsar had been reported before. It achieved the first real-time AI detection of a pulsar, a rapidly spinning stellar remnant that pulses like a cosmic lighthouse, in the Crab Nebula.
Perspective
The demonstrated scope is this agentic observation pipeline at the Allen Telescope Array and one real-time AI detection of a pulsar in the Crab Nebula; the team is evaluating open models including Nemotron for cases where data locality and turnaround matter most and for observatories that may someday run offline on the far side of the moon.
The material describes agents compressing what would take humans weeks or years into hours but gives no corresponding quantitative comparison, and the evaluation of open models is still under way with no results reported.