How SETI astronomers are using AI to finally find ET

Artificial intelligence isn’t just taking jobs away from artists, screenwriters, journalists and office workers. It’s coming for astronomers too – particularly the search for extraterrestrial intelligence, SETI.

However, as with other jobs, experts warn that we should abandon this work AI at our risk. “Nothing can match our insatiable human curiosity and imagination,” Ross Davis, an information and communication sciences specialist at Indiana University, told The Daily Beast.

The threat that self-improving AI chatbots like ChatGPT pose to screenwriters was a driving factor in the six-month strike by the Writers Guild of America, which ended last week with major concessions from Hollywood studios.

The dynamic is different in astronomy, where scientists combine high-level conceptual thinking – i.e. imagination – with rigorous data collection and analysis. The former is a fundamentally human endeavor and should remain so, scientists told The Daily Beast. The latter is already highly automated and could even become so more automated as intelligent algorithms spread across the sciences.

Many astronomers would even welcome more AI support. Andrew Siemion, the director of the SETI Research Center at the University of California, Berkeley, told The Daily Beast that his team is trying to make more of its data — particularly from the powerful Green Bank Telescope — available to AI developers.

In anticipation of the development of increasingly intelligent AIs, a team of SETI researchers led by John Hoang of the Santa Cruz Institute for Particle Physics proposed a division of labor. Algorithms can encode and decode data “at high speed,” Hoang and his co-authors noted a study This appeared online on August 26th. But asking AI to do a nuanced analysis sometimes produces “nonsensical” results, they warned.

This is a phenomenon known as hallucination, or when models like OpenAI’s ChatGPT make up facts. It is an ongoing issue in the field of AI that poses a risk for the spread of misinformation.

Hoang and several of his teammates declined to comment on their study, which did not appear to be peer-reviewed.

In SETI, astronomers typically aim powerful radio receivers at a distant star and record the resulting electromagnetic noise. The goal is to find, within the noise, any broadcasts that may be from an extraterrestrial civilization.

Radio SETI produces a lot of data. In the first decades of these fields – the 1960s and 1970s – astronomers scanned the data in the form of thick stacks of printouts, each page a chaotic numerical recreation of interstellar sounds. They were looking for outliers in the numbers – signs that someone, somewhere, might have said something in the form of a targeted radio broadcast.

It was a breathtaking piece of work that required an equally good mix of number crunching and imagination. After all, an astronomer would have to imagine how aliens might communicate before listening for that communication among the radioactive shrieks and groans of millions of stars being born, growing, and dying.

In one famous case in 1977, astronomer Jerry Ehman noticed, in a stack of printouts showing images from Ohio State University’s powerful Big Ear radio telescope, the contours of a signal that appeared to be 30 times louder than the surrounding stellar noise.

Equally compelling was that the signal propagated at a seemingly symbolic frequency: 1,420 megahertz, the resonance frequency of an energetic hydrogen atom. If aliens were trying to make contact with another, distant civilization, they might first describe something as basic as hydrogen. Something every civilization should recognize.

“Wow!” Ehman wrote in his moment of discovery. His exclamation gave the signal its name.

For several years it has been Wow! The signal was SETI’s great hope: the best chance yet of locating and possibly communicating with an alien civilization. But astronomers were never able to reproduce Ehman’s results. They never heard the same loud signal a second time.

Every few decades or so, a hopeful astronomer organizes a new search for the Wow! Signal. No luck so far. Lately, Columbia University astronomer David Kipping has been making the case for further searches for the signal—and that’s exactly what AI could help by speeding up the analysis of hard drives full of radio data. “Generative AI has the potential to significantly improve the efficiency and effectiveness of the search for extraterrestrial intelligence,” Hoang’s team wrote.

But Ehman’s experience with Wow! All those decades ago is not an argument for AI to take over the search for extraterrestrial intelligence. Ultimately, Ehman’s breakthrough was not analytical, but rather creative. First he noticed a loud signal and then he noticed that the signal seemed to have some meaning. Ehman sensed intent in mimicking the signal of an energetic hydrogen atom. Intelligence. Curiosity. Hope.

These are ideas, not data. Unless the AI ​​is truly sentient, it may never come up with such ideas.

For this reason, the possibilities of a chatbot for astronomy in general and SETI in particular are limited. AI can help overworked and underfunded scientists collect and analyze data. It cannot help them imagine the questions we should answer to better understand the universe. And AI “will not be able to step back and apply intuition to visual elements at a high conceptual level,” Davis said.

Maybe screenwriting, journalism, and office work will run into trouble in the coming AI revolution—but astronomy will be safe for now. When it comes to searching for life in the universe, no bot can yet compete with its flesh-and-blood counterpart. “I don’t think there’s a high risk of SETI,” Annie Robin, an astrophysics researcher at the Institut Utinam in France, told The Daily Beast.

If we left astronomy to the bots, we would turn astronomy into something it is not. We would turn it into a lifeless and pointless exercise in number crunching – rather than a search for understanding as it is supposed to be. “It all starts and ends with us,” Davis said, “with AI being primarily a means rather than an end.”