[ad_1]

“It feels naïve to believe we’re truly alone in the universe”: Meet the U of T student using AI to look for aliens

Peter Ma developed an algorithm to detect extraterrestrial intelligence when he was in high school. Now, he’s continuing the search

When Peter Ma’s high school classmates were supposed to pair up for a computer science project, no one wanted to be on his alien detection squad. The Grade 12 student was planning to look for signs of intelligent life using an algorithm he’d written based on open-source data from the University of California, Berkeley. So Ma—a restless autodidact who had taught himself to code the year before—worked on the project by himself.

Three years later, Ma has no problem finding collaborators. Now studying math and physics at the University of Toronto, he went from being a solo alien tracker to the youngest member of an international team of scientists dedicated to the search for extraterrestrial intelligence (SETI). Using Ma’s high school algorithm, the team has detected eight radio signals that may have originated from an alien civilization. Their findings were published last month in the journal Nature Astronomy, with Ma as the lead author.


How did you become interested in computer science?
I took a class in Grade 11, but it was going really slow. My teacher had all the assignments for the next four months in one folder, so I just did them all in two weeks, teaching myself what I needed as I went. Then I had a lot of free time, so I studied other things I was interested in, like machine learning. When Covid hit, I really had a lot of free time. That’s when I started looking for astronomy projects, because I’d always been interested in that too.

Related: Meet the Etobicoke-born inventor of the ChatGPT detector

You grew up in Markham. When did you start looking at the stars?
When my parents bought me a small telescope, like a lot of people in my field. I was five. I didn’t know what I was looking at—everything, I guess—but it looked cool.

How did you go from class keener to working with UC Berkeley and Breakthrough Listen, a scientific research program searching for civilizations beyond Earth?
In high school, I found Berkeley’s SETI open-source data. They had a call-out for machine learning collaborations on their GitHub repositories, and that’s how I first realized the potential of machine learning in SETI. As I worked on my project, I cold-emailed half the people in the Berkeley SETI lab asking questions about the data and their approach. They were quite appreciative of a relatively young kid being interested and curious, and they eventually offered me a summer job.

How did you spend your summer?
I worked with a small team of undergrads on scaling their search algorithms to cloud computing services, which at that time was relatively unheard of. The goal was to expand search capabilities by moving computing demand away from on-premise resources—our own computers—and onto cloud services like Google Cloud.

How exactly does one go about tracking aliens?
When looking for intelligent life, we can’t actually look for intelligence per se. We look for signs of intelligence—signs of technology. If there’s something out there that’s engineered, something must have engineered it, right? Radio transmitters are pervasive in all the technology we use. Cellphones, wireless, TV satellites—they all emit this kind of electromagnetic radiation, so we look for radio waves using radio telescopes. When we analyze the data, we look for signals that will be small and thin—what we call narrowbands—because signals from interference or astrophysical events like black holes will be super thick.

And how do you tell apart signals from potential alien tech and, say, earthling tech or a black hole?
You start by pointing your telescope at a star. If you see a narrowband signal, that’s good. If you point your telescope away from the star and still see that signal, that’s bad. That means the signal is coming from your environment. So you can look for an on-off pattern. You move your telescope back and forth across a star. If the signal isn’t background noise, it should be there and then disappear.

That’s easy enough to describe, and you could probably identify the pattern yourself right now without much training. But it’s hard to get a computer to do that. The signals could be any kind of shape—a squiggle, a straight line, anything. That’s the problem I’m trying to solve with my algorithm. We teach the computer what to expect by feeding it a bunch of data and then ask it to flag anything that’s anomalous. It can be any shape, as long as it’s unexpected.

Did you find anything unexpected?
We searched through upward of a billion signals and found eight promising extraterrestrial intelligence signals of interest that researchers hadn’t previously identified. But the most significant finding was that deep learning algorithms can successfully perform SETI searches and produce novel results that traditional algorithms didn’t find.

Does that mean you found eight alien cellphones?
Well, the data we were using was collected in 2016 and 2017. When we looked again in 2022, we didn’t find the signals. So we can’t say for sure—the search continues.

What’s next for your project?
Breakthrough Listen’s big goal is to search the nearest million stars for radio signals. We only searched through 1,000 for this paper. We’re working on scaling the algorithm I proposed—which currently works for one telescope—so that it can work on 64 telescopes at the same time.

We’ve seen a lot of unidentified flying objects in the sky and on the news recently. What’s your take on UFOs?
To me, UFOs and extraterrestrial life are two completely different things. There are too many unanswered questions, and we need a lot more data before we can say anything about UFOs. Regarding extraterrestrial life, the more we stare out into the universe and try to understand it, the more we discover that our situation isn’t that unique. Earth is actually exceptionally unexceptional. It feels rather naïve to believe that we are truly alone in the universe. But I’m not going to make any conclusions until I have hard proof.

What kind of proof?
It’s difficult to say. Some researchers at Breakthrough Listen established a checklist for how to evaluate a signal of interest. Finding something that checks all those boxes would be pretty concrete.


This interview has been edited for length and clarity.



[ad_2]

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *