AI Joins Hunt for ET: Study Finds 8 Potential Alien Signals

That also reminds me of Her (2013) and the departure of the AI's, leaving the humans behind. That would be a rather depressing outcome as a human, AI finding alien life but leaving us all behind.

Assuming AGI existed, detected it, and understood the signal’s message — things might get interesting depending on what the message said and how the AGI responded.

ai that talks to other ais grown on other bio substrate

From Accelerando... (talking about an alien signal)

"Colorless green ideas sleep furiously," she suggests.

"Nope," replies the cat. "It was more like: 'Greetings, earthlings, compile me on your leader.'"

...

The cat yawns. "I could have told Pierre instead." Aineko glances at Amber, sees her thunderous expression, and hastily changes the subject: "The solution was intuitively obvious, just not to humans. You're so verbal." Lifting a hind paw, she scratches behind her left ear for a moment then pauses, foot waving absentmindedly. "Besides, the CETI team was searching under the street lights while I was sniffing around in the grass. They kept trying to find primes; when that didn't work, they started trying to breed a Turing machine that would run it without immediately halting." Aineko lowers her paw daintily. "None of them tried treating it as a map of a connectionist system based on the only terrestrial components anyone had ever beamed out into deep space. Except me. But then, your mother had a hand in my wetware, too."

wait what. I just read neuromancer and don't see how that applies? Were rio/wintermute ever involved in detecting an alien signal?

Well, they can train the model on things we know _aren't_ alien signals.

Edit: From the article:

  > To train the AI system, Ma inserted simulated signals into actual data, allowing the autoencoder to learn what to look for. Then the researchers fed the AI more than 150 terabytes of data from 480 observing hours at the Green Bank Telescope.

The simulated signals were sounds that couldn't possibly be produced by any (known) natural astrophysical process. So, the model finds signals that have characteristics of artificially generated signals.

Presumably, because it filtered through orders of magnitude more than that.

Because it’s not so smart, it just try to recognize a set of patterns that was trained to look at

AI is not a well-defined term. This is more likely a specific machine learning technique that is designed to identify all the boring "normal stuff" and ignore it, and then do that at very large scale. By doing that, you can find the interesting parts so that humans with more limited resources can determine if those newly flagged things need to be added to the boring list, or if they do represent something truly interesting.

edit: After reading the article a bit more, it is using a random forest classifier. This is almost certainly not meeting the definition of what many here are thinking when the term "AI" is being used in the title. The term is clearly used here for marketing purposes.

No one claimed the AI "is so smart" - it's another tool in their kit. The article clearly conveys that the potential is a useful way to augment human-led research by providing a way of quickly processing large datasets i.e. the 20k signals are filtering down 150 terabytes of data which is much more manageable for human analysis. It's not like we have definitive parameters of what constitutes an "alien signal," so we can't exactly create an absolute model for detecting such telemetry. Instead the goal of the article is to simply demonstrate exciting new ways to leverage machine learning methods in different contexts

Trying to filter noise from data also follows the 80,/20 principle. The rules for the easy cases is written easily but then you have one-offs, maybes and special cases. Trying to filter one of these does not cover anything else. So doing it manually takes the same time.

Case study: had to clean a list of German addresses once. Excluding obviously invalid addresses like some Chinese address was easy. But some addresses had errors which needed a human eye to fix and correct.

I'm assuming that the 20,515 signals of interest come from a pool that's one or several orders of magnitude larger than 20,515.

Because it isn't AI. It is just statistical models. AI is for marketing.

How do you judge whether 20,515 is a large number or a small number? If the AI provided 2 candidates is that still too high?

Because the statistical models powering AI don't have the depth of understanding of a PhD in this field, and so it casts a wide net to ensure nothing gets missed and then a more refined search by human heuristics is needed.

The others were "intelligent" signals but from human interference. There's a lot of this so the unique patterns don't occur often enough to automate.

The paper is actually worth reading about this part[1]. They have a moderately complex pipeline that you can think of as a filter: it tries to find anomalous signals.

The stage-1 autoregressor filtered 115M signals to 3M. Then they perform signal processing techniques to remove things like GPS signal contamination ("t can be seen that certain observing frequencies contain a much higher number of events compared with the others—for example, the region around 1,600 MHz. This overlaps with known RFI at the GBT site specifically from persistent GPS signals.").

After this second stage they are left with 20,515 potential signals which were visually inspected.

The issue here is that we don't know what a alien signal looks like so we can't just use a classifier. The pipeline can only find things it has never seen before, but it takes human judgement to decide if these signals are "alien" or more likely contamination from human sources that weren't filtered ("Regarding the nature of the rest of the events, most of them look like false positives associated with RFI signals.")

[1] https://www.nature.com/articles/s41550-022-01872-z.epdf?shar...

I think it's useful to replace "AI" with "app" any time you encounter it in the wild. AI is a set of computational techniques that is very broad. The problem with considering "the AI" or "an AI" is that people have in mind some kind of AI agent and that isn't really the case.

AI is also not well defined - day one of my AI course in university opened with "What is AI?" Generally once we figure out how to do something using a computer, we decide that's not really intelligent anymore so the implementation isn't AI. An example of that is the minimax algorithm - it's featured "AI: A Modern Approach"[0] but it isn't really something people think of when they hear "an AI".

[0] https://aima.cs.berkeley.edu/contents.html

The AI is fast, not smart.

DO NOT ANSWER! DO NOT ANSWER! DO NOT ANSWER!

no.

It seems unlikely any alien malware could be targeted to affect our computing platforms. However, if they had preexisting knowledge of computing as implemented on Earth, it would be less complex. If there was a type of malware universal enough to do so, it would probably be some kind of attack against the universal notion of intelligence itself. In which case, you may need to isolate it from humans as well.

I think the most likely scenario for receiving any kind of alien Trojan horse signal, would be if the signal was some kind of instructions for how to create and execute an alien AI as encoded in the signal. However, that would require complex analysis and human intervention to build. Unless we reach a point where AGI systems can search, interpret, and implement instructions from said signals, any threats would require significant involvement from humans in order to materialize.

At least, those are my initial thoughts.

Ha! Have you read Peter Watts's Blindsight and Echopraxia?

No. For now the algorithms that search for extraterrestrial signals aren't trying to decode signals at all. They are only trying to detect extraterrestrial signals. The typical resolution for the Green Bank data is something like, every 20 seconds at a particular frequency gets summarized as one floating point number. That's enough to detect a sufficiently powerful signal over a few minutes of recording, but it isn't enough to decode any message.

If we ever do detect a signal, it will probably require constructing a radio telescope that is even more powerful in order to decode that signal. At that point it would make sense to think about safety measures.

An alien signal for which the act of perceiving it causes some paradox that results in the perceiver to have never existed, cascading to every other interaction they've ever had (not had).

It's a clean and efficient way to nip upstart intelligences in the bud, before they advance far enough to start pulling at loose threads of the fabric of reality (which would threaten the remaining older intelligences, who survived and learned from previous such incidents).

They can do malware given they know how our computers worked and means visited us and could do a lot more than make our machines mine bitcoins

It's a bit of a science fiction myth that it is impossible to process a bit of data without executing it, or that data can somehow force itself to be executed by its sheer swaggering potency as data. And a thing that is more true than it should be, because people keep insisting that they are super awesome and totes capable of writing C that can be put on the network. If we'd stop doing this there'd be hardly any truth to it.

It's not actually that hard to process data without executing it.

Regardless, it is certainly science fiction that they could guess an exploit from light years away. That's not what exploits do, it's not how they work. I've never seen anything like a "universal" exploit that you could just fire at anything, even a non-human system, and have some sort of reasonable expectation that it would work. Such a thing is not even something you could sketch out. Even if you think you have something, like, say, https://github.com/payloadbox/xss-payload-list , you're looking at a human list. All I'd have to do to completely scramble that entire list is to have a parallel evolution of ASCII where the letters and symbols are in completely different places than they are now. Nothing in that list would work if all the control characters in ParallelASCII were in 223-255, and the alphabet was 0-52, and all the symbols were from 128 on. And that's still a very human standard that is, for instance, based on bytes instead of, say, collections of 9 trits as the base level of the system. There's an effective infinity of other ways of encoding things and deciding what characters have what characters doing what things... assuming "characters" is even the way to look at the representation in the first place.

As others mention, you could hypothetically send a program that does something that can't be analyzed, through sheer size if nothing else, but it would still be an uphill battle to just guess how to exploit something. You'd be looking more at an AI that is good enough to talk itself out of the box, rather than something that is actually "hacking" anything reliably.

We should be fine unless we open an alien attachment

My sensors have detected an Excession enjoyer

Not sure how this could even theoretically work. Without having a partial understanding of the intended recipient, a radio signal can't cause harm. Something like 'this signal is false' and 'divide by 0' need to be decoded and processed before they are malicious. Meat space is excellent at detecting and breaking these logic problems and infinite loops. We're safe from evil alien signals. Perhaps a decompression bomb within the signal could cause lots of decoding grief. Now that I think about it, I can imagine us being stupid enough to dedicate all our resources into decompressing or cracking an alien signal to up to the point of extinction. And maybe that's when they step in and call it. We were just part of a great galactic prank.

How improbable is it to imagine a society that doesn't need encryption?

Clearly not happening on Earth, but hardly impossible elsewhere.

> What is the likelihood that intelligent life develops encryption before radio communication

We developed encryption before radio. It just wasn’t very good encryption by today’s standards. The problem is that your best encryption might be trivial to crack by a civilisation that’s thousands of years older than yours.

You think a child would know off the top of their head that the first picture of an exoplanet wasn't taken by the new shiny telescope in the news and not the European Southern Observatory’s Very Large Telescope in 2004?

That it's wrong about this isn't a problem because it's a basic fact everyone knows, but because it gets thing that people don't know wrong in a confident way.

People will assume this answer is as good as looking it up themselves and never check it.

Couldn't you train it to find things that don't correlate to anything from a given dataset that also pattern to non-random/unnatural signal?

"To be sure, because they don’t have real signals from an extraterrestrial civilization, the researchers had to rely on simulated signals to train their models. The researchers note that this could lead to the AI system learning artifacts that aren’t there.

Still, Cherry Ng, one of the paper’s co-authors, points out the team has a good idea of what to look for.

“A classic example of human-generated technology from space that we have detected is the Voyager,” said Ng, who studies fast radio bursts and pulsars, and is currently affiliated with the French National Centre for Scientific Research, known as CNRS.

“Peter’s machine learning algorithm is able to generate these signals that the aliens may or may not have sent,” she said."

They just asked ChatGPT to look /s

> AI? What was the training data? Would seem a bit thin on the ground.

It's in the article.

To train the AI system, Ma inserted simulated signals into actual data, allowing the autoencoder to learn what to look for. Then the researchers fed the AI more than 150 terabytes of data from 480 observing hours at the Green Bank Telescope.

The AI identified 20,515 signals of interest, which the researchers had to inspect manually. Of those, eight had the characteristics of technosignatures and couldn’t be attributed to radio interference.

The researchers then returned to the telescope to look at systems from which all eight signals originated but couldn’t re-detect them.