Artificial intelligence is speeding up astronomy

• AI and machine learning can be used to quickly pick out intriguing parts of a dataset, making it less likely that astronomers will miss something important.
• "With astronomy and these huge datasets, there's always a concern that you missed something, or that you didn't know enough about the objects you're trying to study," astronomer Chris Impey of the University of Arizona tells Axios.

What's happening: Last month, scientists revealed an image of the black hole at the center of the galaxy M87 that was reprocessed using a machine learning algorithm. The sharpened image should allow scientists to more accurately estimate the black hole's mass.

• Scientists have also used artificial intelligence to make it easier to analyze vast amounts of data gathered by gravitational wave detectors like LIGO, picking out the ripples in space and time created by collisions between black holes or other dense objects.
• Now, researchers are using algorithms to figure out characteristics of the objects that produced the gravitational waves in the first place.
• AI is also being used to pinpoint newly forming planets around young stars.

The intrigue: SETI (the Search for Extraterrestrial Intelligence) efforts could be helped by AI and machine learning, which are particularly adept at picking out patterns in large sets of data.

• SETI — which focuses on trying to pick up radio signals emitted by technically advanced societies — generates huge amounts of data.
• "The datasets for SETI endeavors are really massive," SETI Institute CEO Bill Diamond tells Axios. "We generate many tens of terabytes a day."
• The institute and other organizations announced last week that the Very Large Array in New Mexico will join a new experiment to search for radio signals emitted by advanced alien societies, which will rely on machine learning tools.
• "When the compute is completely outfitted for that project — which is right now about halfway — we'll be generating seven terabytes per second of data, so extraordinarily huge amounts of data," Diamond said. "And the only way we're going to be able to parse that data to look for interesting phenomena is with machine learning techniques."

Yes, but: Some worry these tools could spit out false positives that wouldn't otherwise be an issue if humans were analyzing the data.

• "It's possible [with] these huge datasets that it could throw so many candidate anomalies at you that you just couldn't keep up and you end up being buried in the candidate anomalies rather than finding new phenomena," Impey said.
• NASA has also been taking a close look at AI, but a report published last week by the Office of Inspector General stresses there are risks to wide adoption of these tools, including possible cybersecurity threats.

What to watch: Scientists are already training an AI algorithm to help create sharper photos when the Vera Rubin Observatory — tasked with learning more about the nature of dark matter and other science goals — comes online in the coming years.

• Machine learning has been used to find streaks created by satellites streaking overhead in Hubble Space Telescope images, and in the future, algorithms could be used to remove satellite streaks from telescope photos as well.