SETI@home: How Millions of PCs Hunted for Alien Life

<v Speaker 1>Welcome to Bedtime Astronomy. Explore the wonders of the cosmos
<v Speaker 1>with our soothing Bedtime Astronomie podcast. Each episode offers a
<v Speaker 1>gentle journey through the stars, planets, and beyond, perfect for
<v Speaker 1>unwinding after a long day. Let's travel through the mysteries
<v Speaker 1>of the universe as you drift off into a peaceful
<v Speaker 1>slumber under the night sky.
<v Speaker 2>When you really stop and think about it, the search
<v Speaker 2>for extraterrestrial intelligence, it has this incredible romantic quality to it.
<v Speaker 2>It feels like the ultimate human quest.
<v Speaker 3>It does. But there's a flip side, a very practical,
<v Speaker 3>almost brutal reality to it, which is which is that
<v Speaker 3>finding proof, finding that one single artificial signal and all
<v Speaker 3>the noise. It's not a quest for poets. It's a
<v Speaker 3>colossal data processing problem.
<v Speaker 2>Right, It's a triumph of computation over poetry. As you said,
<v Speaker 2>the adversary isn't the distance, not really, it's the sheer
<v Speaker 2>volume of static you have to sift through.
<v Speaker 3>Exactly, and not just cosmic static, our own noise, the
<v Speaker 3>stuff we generate here on Earth is just overwhelming.
<v Speaker 2>And that specific challenge, the problem of computation at an
<v Speaker 2>almost unimaginable scale. That is exactly what we're exploring today.
<v Speaker 3>We're looking at the final comprehensive analysis of the Search
<v Speaker 3>for Extraterrestrial Intelligence at Home project. You know, SETI at Home.
<v Speaker 2>A project that was so much more than a science experiment.
<v Speaker 2>It was a cultural phenomenon. It ran for what twenty
<v Speaker 2>one years, from nineteen ninety nine all the way to
<v Speaker 2>twenty twenty, and.
<v Speaker 3>It was all built on the spare processing power of
<v Speaker 3>millions of ordinary people's computers around the world.
<v Speaker 2>So these final papers, published just last year, they're not
<v Speaker 2>just a simple we looked and found nothing report.
<v Speaker 3>Not at all. For you listening, this is a masterclass
<v Speaker 3>in scientific methodology. We're going to get a shortcut through
<v Speaker 3>two decades of trial and error.
<v Speaker 2>We'll see how they managed to take twelve billion that
<v Speaker 2>number is just staggering, twelve billion raw detections and filter
<v Speaker 2>them all the way down to just one hundred interesting candidates.
<v Speaker 3>And that filtering, that process of winnowing down the data,
<v Speaker 3>that's the real scientific breakthrough here. The lessons they learned
<v Speaker 3>are now foundational for any big sky survey.
<v Speaker 2>Okay, so let's unpack it. The history, the scale and
<v Speaker 2>how crowdsourcing was the only way they could possibly have
<v Speaker 2>solved this.
<v Speaker 3>And I'm ready to get into the technical side of it.
<v Speaker 3>Those final papers, Yeah, how they used these ingenious fake
<v Speaker 3>signals they called them birdies to test their own system.
<v Speaker 3>It's a fantastic story of technical ingenuity meeting just overwhelming scale.
<v Speaker 2>So to really get it, you have to picture the
<v Speaker 2>mid nineteen nineties. It's easy to forget how limited computing
<v Speaker 2>was back then.
<v Speaker 3>Oh absolutely, especially for academic projects. You didn't just get
<v Speaker 3>time on a supercomputer that was for national labs, huge
<v Speaker 3>corporate research, not.
<v Speaker 2>For a bunch of astronomers at a university looking for
<v Speaker 2>aliens exactly.
<v Speaker 3>But that scarcity, it really drove the innovation. Here. Over
<v Speaker 3>at UC Berkeley, you had this computer scientist David Anderson,
<v Speaker 3>and he was already deep into this idea of distributing computing.
<v Speaker 2>Which is basically taking a massive problem and breaking it
<v Speaker 2>into tiny little pieces.
<v Speaker 3>Tiny manageable chunks that you could then send out to
<v Speaker 3>thousands of smaller, less powerful computers to solve independently.
<v Speaker 2>And he just needed the perfect problem to test this
<v Speaker 2>idea on a huge scale.
<v Speaker 3>And the problem found him. He came from one of
<v Speaker 3>his former students David Getty, who suggested why not apply
<v Speaker 3>this to SETI a perfect mask, a perfect marriage of
<v Speaker 3>problem and solution, because analyzing radio data is what we
<v Speaker 3>call an embarrassingly parallel.
<v Speaker 2>Problem, meaning you can analyze one little slice of the
<v Speaker 2>data without needing to know anything about the slice next
<v Speaker 2>to it. They're all independent jobs.
<v Speaker 3>Precisely, you could take a tiny recording from a radio telescope,
<v Speaker 3>chop it up into thousands of different frequency chunks, and
<v Speaker 3>send each chunk to a different volunteers computer.
<v Speaker 2>So they realized they could tap into this huge, growing
<v Speaker 2>network of home PCs, the old Windows ninety five boxes
<v Speaker 2>sitting on people's desks, the.
<v Speaker 3>Ones that were just sitting idle most of the day,
<v Speaker 3>you know, while you're at work or getting lunch. Their
<v Speaker 3>CPUs were just waiting for something to do.
<v Speaker 2>And that was the core idea use that volunteer power
<v Speaker 2>to hunt for artificial patterns, these techno signatures in the noise.
<v Speaker 3>So Anderson, along with Eric Corpela and Dan Werthermer, they
<v Speaker 3>launched it in nineteen ninety nine, and the public response
<v Speaker 3>it wasn't just big, it was seismic.
<v Speaker 2>They had a goal in mind, right, a number of
<v Speaker 2>volunteers that needed to make the science viable.
<v Speaker 3>They did. Their initial calculation was that if they could
<v Speaker 3>get fifty thousand volunteers five zero, that would be enough
<v Speaker 3>computing power to do something genuinely new, something they could
<v Speaker 3>never get a traditional grant for.
<v Speaker 2>Fifty thousand was the benchmark for success.
<v Speaker 3>And what happened was, as Anderson put it, way way
<v Speaker 3>way beyond our initial expectations.
<v Speaker 2>So what were the actual numbers?
<v Speaker 3>Within just a few days of launch, they had two
<v Speaker 3>hundred thousand people sign up from over one hundred different countries.
<v Speaker 2>Wow.
<v Speaker 3>And by the end of that first year they had
<v Speaker 3>two million users, million people donating their spare computer time
<v Speaker 3>to the search.
<v Speaker 2>That's just it's a testament to the public's hunger to
<v Speaker 2>be part of something this profound. It proved the concept
<v Speaker 2>really that you could do real sophisticated science this way.
<v Speaker 3>It absolutely did. It became one of the most visible
<v Speaker 3>and successful citizen science projects ever.
<v Speaker 2>Okay, so we have the computer network, millions of processors
<v Speaker 2>ready and waiting, but what are they processing? Where did
<v Speaker 2>the actual data come from?
<v Speaker 3>It came from a legend the Aricibo Observatory in Puerto Rico.
<v Speaker 2>The giant three hundred meters dish built into a sinkhole,
<v Speaker 2>the biggest in the world. For so long, that.
<v Speaker 3>Was the fire hose. It was the input for the
<v Speaker 3>whole system, just funneling raw cosmic static to Berkeley to
<v Speaker 3>be chopped up and sent out.
<v Speaker 2>And of course that telescope was tragically lost in twenty twenty,
<v Speaker 2>which is what brought the data collection part of the
<v Speaker 2>project to an end.
<v Speaker 3>A huge loss for science. But for twenty one years
<v Speaker 3>it performed flawlessly for this project, and.
<v Speaker 2>The way they collected the data was also key to
<v Speaker 2>why this whole thing was even possible. This idea of
<v Speaker 2>commensal observing right.
<v Speaker 3>Commensalism, it's a term from biology. You know where one
<v Speaker 3>organism benefits and the other isn't really affected.
<v Speaker 2>So in astronomy it means what you're essentially hitching a ride.
<v Speaker 3>You are you're getting data without having to fight for
<v Speaker 3>precious expensive telescope time. See other astronomers would be using
<v Speaker 3>a recibo for their own.
<v Speaker 2>Research, like mapping hydrogen gas in the galaxy or timing pulsars,
<v Speaker 2>things like.
<v Speaker 3>That exactly, and the SETI at Home team just had
<v Speaker 3>their own recording equipment hooked up passively listening in on
<v Speaker 3>whatever the main observer was.
<v Speaker 2>Looking at, so they weren't booking time on the telescope.
<v Speaker 2>They were just recording whatever came in while someone else
<v Speaker 2>was using it.
<v Speaker 3>Incredibly efficient and it's what allowed them to cover so
<v Speaker 3>much of the sky.
<v Speaker 2>How much did they cover over those two decades.
<v Speaker 3>Over the project's lifetime, they observed about a third of
<v Speaker 3>the entire celestial sphere, basically everything you can see from
<v Speaker 3>Puerto Rico.
<v Speaker 2>And critically, it wasn't just a one off look. Repetition
<v Speaker 2>is key, right, it's everything.
<v Speaker 3>You have to see a signal more than once to
<v Speaker 3>believe it. Most of the areas they covered were observed
<v Speaker 3>a dozen or more times, some spots hundreds even thousands
<v Speaker 3>of times.
<v Speaker 2>So, as Anderson said, they basically covered most of the
<v Speaker 2>stars in our galaxy, billions and billions of them, which.
<v Speaker 3>Is why Corpella could confidently say it was without doubt
<v Speaker 3>the most sensitive narrow band search of large portions of
<v Speaker 3>the sky. They had the breath, even if they couldn't
<v Speaker 3>control exactly where the telescope pointed at any given moment.
<v Speaker 2>Okay, so millions of volunteers get their little data packets
<v Speaker 2>from Avacibo. Their computers crunch the numbers and they send
<v Speaker 2>the results back. The input site is solved. Now let's
<v Speaker 2>talk about the output, the result of all that computation, and.
<v Speaker 3>The scale is again just mind boggling. After twenty one
<v Speaker 3>years of this, the collective effort produced twelve billion raw detections.
<v Speaker 2>Twelve billion. Let's define a detection here. What does that
<v Speaker 2>actually mean.
<v Speaker 3>It's just a momentary blip, a little spike of energy
<v Speaker 3>at a particular frequency, coming from a particular spot in
<v Speaker 3>the SAD. The software's job was to flag every single
<v Speaker 3>one of those blips, and for.
<v Speaker 2>Each blip it would analyze frequency, intensity, position.
<v Speaker 3>Right, and to do that used a mathematical tool called
<v Speaker 3>a discrete Fouryer transform.
<v Speaker 2>The DFT. Explain why that's so essential here.
<v Speaker 3>We'll think of the radio signal from space as just
<v Speaker 3>white noise, right, a mess of every frequency mixed together.
<v Speaker 3>A real techno signature an alien beacon would probably be
<v Speaker 3>a very pure, narrow signal, like a single clear note
<v Speaker 3>on a piano, because that's.
<v Speaker 2>The most energy efficient way to be heard over the
<v Speaker 2>background noise. Natural things are noisy and broad, Artificial things
<v Speaker 2>are clean and narrow.
<v Speaker 3>That's the assumption. The DFT acts like a prism for
<v Speaker 3>that noise. It takes the messy signal and breaks it
<v Speaker 3>down into all its component frequencies, into tiny little bins.
<v Speaker 3>So you can spot that one single sharp spike.
<v Speaker 2>But that wasn't the hard part, was it. The real
<v Speaker 2>computational killer was something else.
<v Speaker 3>The real killer was the Doppler drift. This is why
<v Speaker 3>they needed millions of computers.
<v Speaker 2>This is where that incredible ten thousand times multiplier comes from.
<v Speaker 3>It is you have to appreciate all the motion that's happening.
<v Speaker 3>The Earth is spinning, its orbiting the Sun, our whole
<v Speaker 3>solar system is moving, and whatever star system might be
<v Speaker 3>sending the signal is also moving.
<v Speaker 2>It's like the pitch of a train horn changing as
<v Speaker 2>it passes you. A signal scent at a fixed frequency
<v Speaker 2>won't arrive at a fixed frequency because of all that
<v Speaker 2>relative motion.
<v Speaker 3>Precisely its frequency will appear to drift. So if you're
<v Speaker 3>only looking at one specific tiny frequency bin, you'll miss
<v Speaker 3>the signal as it drifts out of that bin.
<v Speaker 2>So the software couldn't just look for a spike. It
<v Speaker 2>had to look for a spike that was moving, that
<v Speaker 2>was drifting across the frequency bins.
<v Speaker 3>Exactly, and since they didn't know how fast the source
<v Speaker 3>was moving, they had to check for all the possibilities,
<v Speaker 3>and all.
<v Speaker 2>The possibilities turned out to be a lot tens of thousands.
<v Speaker 3>For every single data point, the software had to check
<v Speaker 3>tens of thousands of possible drift rates.
<v Speaker 2>So you take your already huge computational problem and you
<v Speaker 2>multiply it by ten thousand.
<v Speaker 3>That's the multiplier. And that's why Anderson says no other
<v Speaker 3>radio SETI project had ever been able to do this.
<v Speaker 3>It was only possible because they could farm out that
<v Speaker 3>insane workload to the public.
<v Speaker 2>So they solved this monumental computing problem. They processed the
<v Speaker 2>twelve billion blips, they did the Doppler analysis, but then
<v Speaker 2>they hit the wall that every radio search hits.
<v Speaker 3>You are a FI wall radio frequency interference, our own junk,
<v Speaker 3>our own junk. It's the central problem in modern SETY.
<v Speaker 3>The vast majority of those twelve billion detections were.
<v Speaker 2>Just We're not just talking about obvious things like TV broadcasts.
<v Speaker 3>Oh no, I mean you have the predictable stuff satellites
<v Speaker 3>in orbit, ground base radar, GPS. But then you have
<v Speaker 3>all kinds of bizarre mundane interference.
<v Speaker 2>Like the famous microwave.
<v Speaker 3>Of an example, the microwady is a classic. The magnetron
<v Speaker 3>inside it pumps out a burst of radio noise. If
<v Speaker 3>a recibo happened to be pointed in just the right
<v Speaker 3>direction to catch a reflection of that off a satellite
<v Speaker 3>or even the atmosphere.
<v Speaker 2>It would look like a powerful narrow band signal from
<v Speaker 2>deep space for a moment.
<v Speaker 3>Yeah yeah, creates a false positive and the real challenge,
<v Speaker 3>as Eric Krpella put it, is the baby with the
<v Speaker 3>bathwater problem.
<v Speaker 2>How do you design your filters to throw out all
<v Speaker 2>the noise without accidentally throwing out the one real signal
<v Speaker 2>you're looking for.
<v Speaker 3>It's a huge methodological challenge. If your filter is too aggressive,
<v Speaker 3>you could delete a genuine, faint signal just because it
<v Speaker 3>happens to look a little bit like a satellite passing overhead.
<v Speaker 3>You have to be able to prove what you're.
<v Speaker 2>Excluding, and this leads us to one of the most
<v Speaker 2>interesting parts of the story, a really candid admission.
<v Speaker 3>From the team the planning flaw.
<v Speaker 2>They had this brilliant solution for the front end problem,
<v Speaker 2>the data collection and processing, but the back end, the
<v Speaker 2>analysis of all those results. It lagged way behind.
<v Speaker 3>Yes, David Anderson was very open about this. He said
<v Speaker 3>that for years, basically up until around twenty sixteen, So
<v Speaker 3>for seventeen years, for seventeen years, they were accumulating all
<v Speaker 3>these detections from volunteers, but they didn't really know what
<v Speaker 3>we were going to do with them. They had fully
<v Speaker 3>worked out the second half of the process.
<v Speaker 2>That's amazing, data is just piling up for over a
<v Speaker 2>decade and a half. Why what was a bottleneck.
<v Speaker 3>Well, the initial priority was just getting the volunteer computing
<v Speaker 3>part to work the BOIANC platform. And remember in the
<v Speaker 3>early days you had people on dial up modems, so
<v Speaker 3>the analysis on the home PC had to be kept
<v Speaker 3>relatively simple. The really heavy lifting cross correlating all twelve
<v Speaker 3>billion detections against each other to find signals that appeared
<v Speaker 3>more than once from the same spot. That required a
<v Speaker 3>massive centralized computer with a ton of memory.
<v Speaker 2>And storage, which they didn't have at the beginning.
<v Speaker 3>They didn't have it funded or available. So the solution
<v Speaker 3>meant a complete shift in strategy from millions of tiny
<v Speaker 3>computers to one giant one.
<v Speaker 2>And that's where the Max Planck Institute in Germany came in.
<v Speaker 3>Exactly, they got access to a powerful computing cluster there,
<v Speaker 3>and that's what finally allowed them to do the big
<v Speaker 3>sophisticated culling process.
<v Speaker 2>So what did that supercomputer do that the home couldn't.
<v Speaker 3>It systematically ran the final stage filtering algorithms across the
<v Speaker 3>entire data set. It was looking for known RFI signatures,
<v Speaker 3>and most importantly, it was checking for persistence over time.
<v Speaker 3>Did a signal from this spot in the sky show
<v Speaker 3>up in two thousand and two, and then again in
<v Speaker 3>two thousand and eight, and again in twenty fifteen.
<v Speaker 2>And that's how they did the first major cut.
<v Speaker 3>That's how they got from twelve billion detections down to
<v Speaker 3>a much more manageable number, a couple of million signal candidates.
<v Speaker 2>So a candidate is a signal that showed up more
<v Speaker 2>than once from roughly the same place, more or less the.
<v Speaker 3>Same place, more or less the same frequency, but spread
<v Speaker 3>out over time. That's the key. That repetition is what
<v Speaker 3>makes it interesting.
<v Speaker 2>Okay, so a couple million candidates. Now they have to
<v Speaker 2>prove their search actually worked. They need to define how
<v Speaker 2>sensitive it was.
<v Speaker 3>And this is where my favorite part of the whole
<v Speaker 3>project comes in. It's just so clever.
<v Speaker 2>The birdies test, the birdie's test. It sounds like something
<v Speaker 2>from golf.
<v Speaker 3>It's one of their greatest scientific legacies. To figure out
<v Speaker 3>how good their system was, they created about three thousand
<v Speaker 3>fake signals, the birdies.
<v Speaker 2>And they injected these fake signals into their own data pipeline.
<v Speaker 3>They did right at the very beginning, before any of
<v Speaker 3>the filtering started, and crucially, they blinded themselves. They didn't
<v Speaker 3>know where the fake signals were or how powerful they were.
<v Speaker 3>They just treated them like any other blip from the telescope.
<v Speaker 2>So it's a perfect self calibration. They run the entire
<v Speaker 2>twenty one year data set through all their filters, and
<v Speaker 2>then at the end they checked to see how many
<v Speaker 2>of their fake birdies survived the process.
<v Speaker 3>Precisely, if they put in one hundred birdies at a
<v Speaker 3>certain power level and only ninety five came out the
<v Speaker 3>other end, they know their system has a ninety five
<v Speaker 3>percent detection efficiency for signals of that power.
<v Speaker 2>Which lets them make a really powerful scientific statement.
<v Speaker 3>The definitive one. They can finally say, if there were
<v Speaker 3>a signal above this specific power level in the parts
<v Speaker 3>of the sky, we looked at we would have found it.
<v Speaker 2>That's huge. So it's not just we didn't find anything,
<v Speaker 2>it's we didn't find anything stronger than x. It says
<v Speaker 2>a hard limit.
<v Speaker 3>For the It sets the baseline for all future searches.
<v Speaker 3>But you know, this rigorous process also forced them to
<v Speaker 3>be very critical of their own work. Anderson admitted the
<v Speaker 3>project didn't completely work the way we thought it was
<v Speaker 3>going to.
<v Speaker 2>What kind of flaws did they find in their own process?
<v Speaker 3>A lot of it came down to compromises they had
<v Speaker 3>to make because of the technology in nineteen ninety nine.
<v Speaker 3>For example, to save processing time on those slow home PCs,
<v Speaker 3>the Doppler driftbins they checked were maybe a little too wide.
<v Speaker 2>So a real signal could have what fallen between the cracks.
<v Speaker 3>It could have if its drift rate was exactly between
<v Speaker 3>two of the rates they were checking, it might have
<v Speaker 3>been missed even if it was strong. Similarly, some of
<v Speaker 3>the early RFI filters were a bit too simple, and
<v Speaker 3>they might have accidentally thrown out a good signal because
<v Speaker 3>it looked too much like a known piece of interference.
<v Speaker 2>All right, so they've got their millions of candidates, they've
<v Speaker 2>tested their sensitivity. Now they have to get to the
<v Speaker 2>final list, the real contenders.
<v Speaker 3>Right. They applied a ranking system to those millions of candidates,
<v Speaker 3>looking at things like how persistent they were, how clean
<v Speaker 3>the signal was, That got them down to the top thousand.
<v Speaker 2>And at this point the algorithms are done, it's time
<v Speaker 2>for the humans to step in.
<v Speaker 3>Corpula and Worthemer they personally manually reviewed those top thousand signals.
<v Speaker 2>What are they looking for at that stage? What makes
<v Speaker 2>one signal better than another?
<v Speaker 3>They're looking for the smoking gun. They're examining the frequency
<v Speaker 3>plots by hand, checking the signals behavior against databases of
<v Speaker 3>known satellite paths, known sources of terrestrial RFI. It's painstaking detective.
<v Speaker 2>Work, and that intense human led review is what produced
<v Speaker 2>the final list, the.
<v Speaker 3>Final one hundred, about one hundred signals that are persistent, narrow,
<v Speaker 3>and that they cannot definitively explain away as our own noise.
<v Speaker 3>These are the ones deemed worthy of a second look.
<v Speaker 2>And the criteria for what's worthy is all based on
<v Speaker 2>this big assumption we have about what an alien signal
<v Speaker 2>would even look like.
<v Speaker 3>Right, Absolutely, the whole search is guided by this model
<v Speaker 3>of an idea deal techno signature.
<v Speaker 2>And that ideal is what a big, powerful narrowband signal.
<v Speaker 3>Powerful narrow band and probably at a special frequency, a
<v Speaker 3>universal signpost.
<v Speaker 2>The famous water hole.
<v Speaker 3>Exactly specifically near the twenty one centimeter wavelength, which is
<v Speaker 3>the frequency of neutral hydrogen gas. Hydrogen is the most
<v Speaker 3>common thing in the universe. Yeah, So the logic is
<v Speaker 3>any advanced species would know to look there. It's like
<v Speaker 3>putting a billboard on the intergalactic highway.
<v Speaker 2>So the assumption is a two part message, a big,
<v Speaker 2>simple hello, and then the actual content.
<v Speaker 3>That's the working hypothesis. You have this powerful narrow band beacon,
<v Speaker 3>it's just a flag just to get our attention. Once
<v Speaker 3>we find that, we'd point everything we have at that
<v Speaker 3>spot and look for a weaker, wider band signal right
<v Speaker 3>next to it that contains the actual information.
<v Speaker 2>But let me push back on that a bit. Aren't
<v Speaker 2>we just projecting our own technology onto them. We're looking
<v Speaker 2>for radio because that's what we know. A civilization millions
<v Speaker 2>of years ahead of us might be using something we
<v Speaker 2>can't even imagine.
<v Speaker 3>That is a fundamental philosophical problem at the heart of
<v Speaker 3>SETI Ye, you're absolutely right, we're looking for something we
<v Speaker 3>would know how to build. It's a search into the street.
<v Speaker 2>Light, so we could be totally deaf to a signal
<v Speaker 2>that's using i don't know, gravity waves or something.
<v Speaker 3>Or neutrinos or some kind of quantum communication. Yes, so
<v Speaker 3>this result, this non detection, is powerful, but it's constrained
<v Speaker 3>by that very specific assumption about the communication.
<v Speaker 2>Method, which brings us to the ultimate conclusion. After all
<v Speaker 2>of that work, the twenty one years, the millions of volunteers,
<v Speaker 2>the supercomputers, the scientific fact is no definitive signal was detected.
<v Speaker 3>And that's a tough pill to swallow. Corporala admitted to
<v Speaker 3>feeling a little disappointment that we didn't see anything in
<v Speaker 3>that one very specific sense, the primary goal wasn't met.
<v Speaker 2>But in every other sense it was a massive success.
<v Speaker 3>A monumental success. Anderson was very clear about this as
<v Speaker 3>a project for community engagement and for proving a new
<v Speaker 3>way of doing science. It went so far beyond their
<v Speaker 3>wildest dreams. It literally pioneered this entire field of volunteer computing.
<v Speaker 2>And they put a definitive cap on it with the
<v Speaker 2>publication of those two big papers last year in the
<v Speaker 2>Astronomical Journal.
<v Speaker 3>Those papers are the project's technical legacy. The first one,
<v Speaker 3>data Acquisition and front end processing, is basically the blueprint
<v Speaker 3>for how to collect and distribute this kind of data, how.
<v Speaker 2>The Aricibo Commensal observing worked, the architecture of the BOIONC
<v Speaker 2>platform right.
<v Speaker 3>And the second paper, data Analysis and Findings, is all
<v Speaker 3>about the back end, the filtering, the RFI rejection, and
<v Speaker 3>crucially the results of that Bertie's test. It lays out
<v Speaker 3>the exact sensitivity they achieved.
<v Speaker 2>And it provides the list of one hundred candidates. It's
<v Speaker 2>not just the result, it's showing all their work so
<v Speaker 2>the entire scientific community can build on it.
<v Speaker 3>That's right. It's a completely transparent, peer reviewed conclusion to
<v Speaker 3>one of the biggest science experiments ever run.
<v Speaker 2>And that's not the end of the story, because that
<v Speaker 2>list of one hundred signals is now the focus of
<v Speaker 2>a new search.
<v Speaker 3>The journey continues. Those one hundred potential golden tickets are
<v Speaker 3>now being targeted by a much much more powerful telescope.
<v Speaker 2>China's five hundred meters aperture spherical telescope fast.
<v Speaker 3>FAST is the logical next step. It's the new king
<v Speaker 3>of radio astronomy, and it has a game changing advantage
<v Speaker 3>over a recibo.
<v Speaker 2>How much more powerful are we talking in.
<v Speaker 3>Terms of raw collecting area, which is what really matters.
<v Speaker 3>Fast is about eight times bigger than a receival was eight.
<v Speaker 2>Times, So that translates directly into sensitivity, right directly.
<v Speaker 3>It means if a signal was truly real but just
<v Speaker 3>too faint for a recivo to see clearly, Fast should
<v Speaker 3>be able to pick it up with ease. You can
<v Speaker 3>see things that are eight times fainter.
<v Speaker 2>So the follow up is already happening.
<v Speaker 3>It is since last summer. Fast has been pointing at
<v Speaker 3>these under targets. For each one, it's staring for about
<v Speaker 3>fifteen minutes, and.
<v Speaker 2>Fifteen minutes with Fast is worth a lot more than
<v Speaker 2>fifteen minutes with a recibo.
<v Speaker 3>A whole lot more. Now, the analysis of that new
<v Speaker 3>data isn't complete yet, Anderson is very measured about it.
<v Speaker 3>He doesn't necessarily expect to find anything.
<v Speaker 2>You have to look.
<v Speaker 3>You absolutely have to look. It's the only way to
<v Speaker 3>close the loop on the SETI at Home project. If
<v Speaker 3>a signal does reappear, it becomes the biggest discovery in
<v Speaker 3>human history. If none do, it makes the original negative
<v Speaker 3>result that much stronger.
<v Speaker 2>This follow up also highlights a big strategic difference in
<v Speaker 2>how we search. SETI at Home was this huge all sky.
<v Speaker 3>Survey, casting the widest possible net.
<v Speaker 2>Whereas most modern projects like Breakthrough Listen are doing targeted searches.
<v Speaker 3>Right they're taking telescopes like Green Bank or Meerkat and
<v Speaker 3>pointing them at very specific nearby stars that we know
<v Speaker 3>have planets.
<v Speaker 2>So they're trading that wide coverage for a much deeper,
<v Speaker 2>more concentrated look at a few promising spots.
<v Speaker 3>And that's the fundamental trade off. Corporala points out that
<v Speaker 3>even with our best modern telescopes, a targeted search can
<v Speaker 3>only detect a transmitter the size of a recibo if
<v Speaker 3>it's relatively nearby.
<v Speaker 2>Galactically speaking, they can't probe the other side of the
<v Speaker 2>galaxy very effectively.
<v Speaker 3>Not really. And the other big difference is control. SETI
<v Speaker 3>at Home couldn't control the telescope. They were just along
<v Speaker 3>for the ride.
<v Speaker 2>Whereas Breakthrough Listen has dedicated telescope time they can point
<v Speaker 2>at where they want for as long as they want.
<v Speaker 3>And Corporal says that's really the ideal scenario. Now, having
<v Speaker 3>controlled is everything, So it's this constant tension in the field.
<v Speaker 3>Do you go for breadth or do you go for depth?
<v Speaker 2>And even though the SETI project itself is over, the
<v Speaker 2>technology it created is more alive.
<v Speaker 3>Than ever the BOIANC platform. That's the lasting engineering triumph.
<v Speaker 3>Anderson didn't just build a piece of software for one project.
<v Speaker 3>He built an open source framework that any massive science
<v Speaker 3>project can now use, and.
<v Speaker 2>They are using it for all sorts of things completely
<v Speaker 2>unrelated to aliens.
<v Speaker 3>The diversity is amazing. You have Rosetta at home, which
<v Speaker 3>is using volunteer computers to calculate how proteins fold. That's
<v Speaker 3>critical for designing new drugs.
<v Speaker 2>There's Einstein at home too.
<v Speaker 3>Searching for pulsars and gravitational wave data. And LHC at
<v Speaker 3>home is helping SERN simulate particle collisions. Legacy is thriving
<v Speaker 3>in biology, astronomy, physics.
<v Speaker 2>So could you do another SETI crowdsourcing project today? Is
<v Speaker 2>it still a viable idea?
<v Speaker 3>Corpela thinks it's absolutely feasible and probably easier in many ways.
<v Speaker 3>Computers are faster, of course, but the really big change
<v Speaker 3>is internet.
<v Speaker 2>Bandwidth, right, no more dial up modems exactly.
<v Speaker 3>They had to keep the data chunks Tony back then. Today,
<v Speaker 3>with broadband, you could send much bigger, more complex chunks
<v Speaker 3>of data to each volunteer, which would be way more efficient.
<v Speaker 2>So the technology is better, the volunteer spirit is probably
<v Speaker 2>still there. What's the roadblock?
<v Speaker 3>Money But not for the computers. For the people.
<v Speaker 2>You still need a paid staff to run the whole thing.
<v Speaker 3>You need engineers, scientists, project managers. As Corpella notes, While
<v Speaker 3>the computing is free, it's not the cheapest way to
<v Speaker 3>do SETI because you have to fund the salaries of
<v Speaker 3>the team coordinating millions of people and billions of data points.
<v Speaker 3>That's the real bottleneck.
<v Speaker 2>Now, as we get to the end of our look
<v Speaker 2>at this, there's one final thought from Repella that I
<v Speaker 2>find really haunting.
<v Speaker 3>It's the great hope and maybe the great frustration of
<v Speaker 3>any archival search. He said, there's still the potential that
<v Speaker 3>et is in that data, and we missed it just
<v Speaker 3>by a hair.
<v Speaker 2>And not because of carelessness, but because of those engineering
<v Speaker 2>compromises they were forced to make back in nineteen ninety
<v Speaker 2>nine exactly.
<v Speaker 3>He said, if he had the funding, his dream would
<v Speaker 3>be to reanalyze the entire twenty one year data set
<v Speaker 3>from scratch to fix the mistakes that.
<v Speaker 2>We made, like using finer toothed combs for the Doppler
<v Speaker 2>crifts so nothing falls between the.
<v Speaker 3>Cracks, and applying all the new sophisticated AI driven RFI
<v Speaker 3>filters that have been developed since a signal that was
<v Speaker 3>wrongly flagged as noise in two thousand and three might
<v Speaker 3>be correctly identified as a candidate today. The data is
<v Speaker 3>still there, it's just waiting for better tools.
<v Speaker 2>The potential for discovery is still sitting on those hard drives.
<v Speaker 2>That's a powerful idea.
<v Speaker 3>So if we were to boil it all down, there
<v Speaker 3>are probably three huge takeaways from this whole incredible saga.
<v Speaker 2>Okay, let's hear them.
<v Speaker 3>First, just the historical scale of it. It proved the
<v Speaker 3>power of crowdsourcing, brought millions of people into the scientific
<v Speaker 3>process and gave us the BOIANC platform that's still driving
<v Speaker 3>science today.
<v Speaker 2>Second, the sheer technical challenge tackling twelve billion signals, dealing
<v Speaker 2>with that ten thousand x Doppler multiplier and building the
<v Speaker 2>filters to get that down to just one hundred candidates.
<v Speaker 3>And third, the scientific legacy using those birdies to establish
<v Speaker 3>a hard scientific sensitivity limit. They gave future astronomers a
<v Speaker 3>definitive line in the sand. We now know what wasn't
<v Speaker 3>out there, which helps us focus on where to look next.
<v Speaker 2>So SETI at Home didn't find ET, but it wrote
<v Speaker 2>the rule book for how to conduct the modern search, which.
<v Speaker 3>Brings us to the final really provocative thought.
<v Speaker 4>What if.
<v Speaker 3>What if one of those one hundred candidates now being
<v Speaker 3>checked by fast actually reappears.
<v Speaker 2>If that powerful narrowband signal is confirmed.
<v Speaker 3>Just once, everything changes, Every telescope on Earth, radio and
<v Speaker 3>optical would turn to that one single spot in the sky.
<v Speaker 2>And you have to consider the impact of just that confirmation,
<v Speaker 2>not the message, not the content, just the simple verified
<v Speaker 2>fact that a beacon is there, The hello, we are here.
<v Speaker 3>Humanity's place in the cosmos would be fundamentally redefined. In
<v Speaker 3>that instant, our cosmic loneliness would be over. And that knowledge,
<v Speaker 3>that possibility is what keeps the search alive.
<v Speaker 2>It's what keeps scientists looking not just at new data,
<v Speaker 2>but back into those old archives, waiting for the right
<v Speaker 2>key to unlock their secrets.
<v Speaker 3>The results are in, but the story might not be over.
<v Speaker 4>Thank you for joining us for this look into the sources.
<v Speaker 4>We'll see you next time.
<v Speaker 3>The us
<v Speaker 4>Sti