We'll Return, After This Message

by John Walker
kelvin@fourmilab.ch
December 1st, 1989

When the foundations of everything you think you know shift beneath you, you can feel it. The night Art Crane and I found the Message, it felt like that moment at the onset of an earthquake when you realize the floor is really moving. Even after twenty years, I can't recall that night without seeing reality shimmer slightly, like the distant mountains on a hot day.

What we found that January night expunged a century's accumulated smugness about our place in the universe. And the funny thing is, we weren't even looking in the right place.

In December of 1997 it seemed as if humanity was well on its way to figuring out the universe. The discovery of the Proteus particle by de Vany, Trang, and Zweig handed the astronomers an all-in-one answer to the riddles of the missing mass and the solar neutrino deficit. The human genome project was winding down having yielded, if little understanding, plenty of data. The Soviet outer planet robots were flying in formation toward their Jupiter gravity kick, thence to Voyager's ports of call and onward to Pluto and Charon. The Shuttle was expected to return to flight any month. That glorious daylight supernova of 1996 was just beginning to fade from the nighttime sky.


I figured Crane might call. When I answered the phone on that rainy day after Christmas, I wasn't surprised to hear his usual request, "Cliff--I'm down here at the office. Could you come in and help me with something?"

Art Crane's a programmer, and a damned good one. But mostly, he is the Bach of the wild-ass conjecture. You can't spend ten minutes with him without hearing him suggest a new mechanism for speciation in biology, opining that the roots of monetary inflation lie in the domestication of animals, or wondering how we'd know if just a few electrons had different masses. He writes every idea down in a little notebook he always carries, then he types them all in to his machine every night. I haven't asked him how big the file is. He reads everything, seems to know all, has opinions on any topic you can name but will gladly argue either side. He brooks no inaccurate facts or sloppy reasoning. He's the kind of person who'd be intimidating and unapproachable, if only he ever finished anything.

He isn't a flake. At least not all the time. When he gets a Big Idea, he's like a terrier. He grabs it and shakes it till it falls apart. Then he sniffs at the pieces. For some reason, whenever he has a Big Idea, I always seem to get involved. Not that I mind. Except for the stock market idea. That one I minded.

Christmas day, Crane and I had been invited to Hack Watkins' for turkey dinner. Later, as we watched Watkins' kids reduce their holiday bounty to pieces siftable through chicken wire, Art was holding forth on his latest idea--second-hand SETI. He was fascinated by pre-discovery observations. Galileo spotted Neptune and even charted it next to Jupiter in one of his notebooks. If the next night hadn't been cloudy, he'd probably have discovered it more than 200 years before Adams and Le Verrier.

The year before, more than a hundred amateur astronomers took pictures showing the Orion supernova brightening before it burst into naked-eye visibility, but not one noticed it till after the fact. Crane said this was inherent in modern science; building big new machines and using them for bold searches was sexy and easy to fund, but rarely did anybody rummage through the dusty archives until something interesting had turned up in new data. He figured that if we ever received a signal from another civilization, we'd probably find dozens of others buried in the archives, easily located once we knew what to look for.

"The facts, dear Clifford, are not in our stars, but on our shelves." he said, "Why don't we look there?" When I headed home around midnight, he and Hack were kicking ideas back and forth about how image processing tools might be used to identify intelligent signals.

I showed up at the office around sunset. Not a soul was there, and only a few programmers. Crane, who never rose before the crack of noon, rarely undertook serious work before six. I was surprised to discover him already in his office, surrounded by a midden of books, pieces of paper, and partially consumed processed food-like substances suggesting, by its height, that he'd been there for several hours. He turned from the screen as I walked in, "Cliff, glad to see 'ya. Look, we gotta get more crunch power on this job."

This was the week for it. The company always closed between Christmas and New Year's. It was a tradition they called the "Annual Week Of Rest," which was a fine joke because there was another tradition that the programmers would use a week devoid of managers, marketeers, meetings, and memoranda to try to out-do one another in huge bursts of concentrated effort to impress each other and just incidentally enrich their employer when they staggered back, bleary eyed, to another year of "regular work." Not that it was expected, of course--but all the computers were left running that week anyway. The Exalted Founder had started it 15 years ago and still kept at it, at least in theory. Nobody could tell, actually; he'd gone through some kind of comprehensional singularity and nobody understood anything he'd done in the last five years.

Back in '97 Xanadu still wasn't finished, but if you knew where to look and how to access it on the Net, you could get most of the machine-readable raw science data since 1970. The Net was installing a new software release over the holidays and had declared connect time and data transfers free between Christmas and January 5th to encourage users to test it. Art proposed to make the most of this. He planned to search all kinds of astronomical data, from the earliest radiotelescope sweeps to the downlink from the gamma ray imaging telescope in that converted shuttle tank with an algorithm he called the "annoyance filter."

"Whatever the message is, and however they encode it, it's going to be obvious, at least in retrospect. Besides," he said, "we have all the inarticulate bozos we need on Earth. There's no need to scour the galaxy for more."

He figured the message would be a picture, and that it would be deliberately made easy to distinguish from a noisy background. This was right down Crane's alley. A couple of years before he'd been obsessed with the idea of developing a program to discriminate television programs from commercials. He wanted to start a company to make boxes that paused VCRs when commercials came on. He considered it an artificial intelligence challenge, "If any idiot can fast forward past a commercial, why can't we design a program to zap 'em?" He spent the better part of a year's spare time tweaking and tuning his algorithm before abandoning it; toward the end it worked pretty well, but not good enough to sell--it recorded moody commercials and edited out climactic scenes of cop shows. It did zap all the car dealer ads.

He likened the problem to protective coloration. "If television is a medium that delivers entertainment at the price of advertising, then advertising and entertainment will co-evolve to become indistinguishable in time." But in SETI, he believed, the incentives were different.

"Signal to noise! Look, are you going to go before the Congress of Galactic Elders Subcommittee on Unessential Projects and try to justify spewing terawatts of soft-sell to the stars? Whatever they're sending, it's going to be obtrusive, blatant, shrill, noisy, coarse, and puffing. It will be calibrated to attract, to rouse, and to entice. Count on it...it's advertising."

He started by taking the commercial zapper and optimizing it for single still frames by training it on magazine advertisements. Then he wrote a front end that scanned bit streams and applied a Fourier transform to recognize scan-line encoded images. These he planned to link into a tool that could process several hundred megabytes of raw data per hour per machine, generating very few false positives for intelligent messages. When I showed up, he said it was "coming along."

There are several distinct phases in an Art Crane project. The first is "inspiration;" he'll be consumed, usually without warning, by a Big Idea. He'll corner everybody in sight, talking a hundred words a minute, filling white boards with diagrams, shoving a sheaf of yellow paper in people's faces, and otherwise explaining why what he just thought of is not only the most important project on the planet at that particular moment, but painfully obvious to any vertebrate. Art has a talent for seeing how tools can fit together to do things they weren't designed for. His ability to estimate the difficulty of all the tasks involved in his plans is more modest; Hack Watkins once said "The only constraint on Crane's armwaving about holes in his designs is that his fingertips can't exceed the speed of light."

"Coming along" means he's making steady progress, but doesn't have much to show for it. Eventually, he'll get an initial, crude version to work, achieving a characteristically grandiose milestone he calls "initial operating capability." Thereupon he invariably "goes ballistic" as he begins to realize all the things he can add to the initial version. "Ballistic" is evocative not only of Crane's disregard of external guidance, but also of the frenzy and round-the-clock concentration that characterizes his efforts, rendering him in this mode more a force of nature than a colleague. This ends at a point of exhaustion when he's run out of things to add to the product or can't sustain the kind of effort he's been putting in, whereupon he enters the "gory details" period as he attempts to clean up all the loose ends and render the result usable to people other than himself. Finally, he deliberately "throttles back," catches up on his reading, and comes within a standard deviation or so of a normal human being as he awaits the next Big Idea.

The progression of phases is as predictable as the Moon's, but their timing has none of the regularity of the cosmos, much to the exasperation of all who work with him. Still, when he is good, he is very good indeed. Three times in the last decade he had singlehandedly come up with the concept and initial prototype of products that now collectively accounted for half the company's sales--each one stemming from a Big Idea unrelated to his regular work. This track record made management more than willing to endure his eccentricities and propensity to indulge in what Watkins called "art for Art's sake."

He picked up the pizza box to the left of his keyboard and carefully stacked it atop a pile of books to clear a workspace. Grabbing a yellow pad and pen, he explained what he needed from me. "We can do this, but we've only got nine days till the Net starts charging again. Remember that cycle sucker animal you built for the inlet problem? That's what we need to drive the filter."

For once, what he wanted wasn't that hard. Last year when we'd worked on a hypersonic fluid flow problem for the Japanese spaceplane, I'd partitioned the job to throw all the idle time on all of our workstations at it. This was easier. All I had to do was take the data stream from the Net and apportion it out to all the compute resources I could lay my mitts on with an eenie-meeine-miney-moe distributor. With 500 workstations rated at a gigaflop or better sitting idle until the new year, we'd be able to cull a lot of data in a few days.

I went down the hall to my office, turned on the monitor, and set to work. The linear flow of real time changed into programming hours, measured more by the accumulation of pop cans, pizza and chinese take-out boxes, and crumpled sheets of yellow paper than numbers on the wall clock.

Art continued to tune the annoyance filter while I developed tools to obtain raw data from the Net, translate them into a uniform format his program could read, and partition the compute job among the machines in the building. Hours passed, then days. I went home and slept when I was too tired to go on, and I presume he did too.

We'd both finished by Monday the 29th. When I arrived, Art had queued me mail saying "Ready when you are." I walked down the hall thinking how odd it was that I'd been collaborating on a project with him, working in the same building, but hadn't seen him for four days. When I reached his office, I realized this was a good thing; the remains bespoke a major ballistic episode. Crane's office resembled a centerfold from Toxic Waste Monthly. I suppose an archaeologist could date the slices of pizza by the quantity of mold. I focused on the task at hand.

It took a couple of hours to integrate Crane's filter with my dispatcher. When we were done, we ran some tests to measure the total compute power we were getting, then some known data checks to make sure we hadn't messed anything up. Around midnight we were ready to start on live data. For the last several days I'd been keeping our Net link busy transferring files we wanted to search to our local server since we could process on-site data much faster than remote files across the Net. Also, I wanted to copy as much data as I could while Net access was free. I didn't have the time to be very discriminating in my acquisitions; if a Net site had some files of interest and didn't publish a huge amount of data, I just figured, "Grab it all. Let Crane sort it out."

I entered the command to start the root task. It cloned itself onto every active workstation in the building, then each copy started pulling files to be searched from the master task queue and commenced scanning them with Crane's annoyance filter. The search was singularly unexciting. My task monitor showed the volume of data we'd scanned so far, the percent done, and the instantaneous compute power we were using. There's a feeling of power that comes from knowing you're doing more calculations every minute than all of mankind did from Pithecanthropus Erectus to 1950, but the novelty of even so remarkable a fact quickly pales into something akin to watching paint dry.

Every time Crane's annoyance filter triggered, my control program popped up a window with the image on both our workstations and filed the picture for later examination in a directory called WOW. The filter seemed to be working quite well. We got an image about every hour--mostly scanned illustrations from journal articles erroneously filed in raw data directories. Art figured we could ditch them by looking for captions, but they were appearing so infrequently he decided not to delay the search to modify the program. Around three A.M. Tuesday we were both taken aback when we found our first genuine interstellar message: a spacecraft, two beings, and some coded gibberish. Art looked at the source information in the title bar and confirmed that we'd discovered the Pioneer plaque in a NASA Goddard planetary image archive. "Right message, wrong way," he muttered. We decided to call it a night. With the search on autopilot, we could dial in from our home machines and scan the images in the WOW directory whenever we wanted. I, for one, had no desire to spend any more time in Crane's office until the janitors came in next week and cleaned it with a fire hose.

Tuesday, I slept late, gave the plants their biennial sip of water, unloaded the dishwasher, and didn't get around to dialing in to see if anything had been found until close to midnight. I looked over the 20-odd pictures in WOW and found the usual selection of false alarms. I checked the login history and saw that Art was dialing in every couple of hours to scan the pictures. The company was closed until Monday the 5th, so we'd have the better part of a week to continue the search. I figured we could get through about a quarter of the obvious candidate data on the Net in that time. The audacity of two programmers employed by a medium-sized software company searching three decades' accumulated scientific knowledge for interstellar messages didn't occur to me. I doubt Crane was capable of entertaining such a thought.

With the search for intelligent life in the universe toiling away, needing only sporadic attention, I turned my attention to serious work. The company had bought one of the just-developed superconductive long range NMR scanners, and I was working on a prototype for a product I called "Fantastic Voyager" that let you fly around in a 3D image of your own body using the cyberspace gear we made. The job was challenging both from the standpoint of identifying the different tissue types from the NMR samples, and in developing navigation models to let you trace the paths of blood vessels and nerves through the body. Since my home machine had all the power and storage I needed, I worked there, as usual disappearing into the project and surfacing only sporadically to eat, sleep, or dial in to scan the WOW directory.

Wednesday night I stopped by Bart Lazslo's New Year's Eve party for a couple of hours and was amused to hear Crane grandiloquently describing our little holiday hack as the "Crane/Slatkin Quest For Galactic Intelligence." He related our discovery of the Pioneer plaque much more dramatically than I could have, or, for that matter, than the facts justified. He'd brought a portable computer and hooked it up to show pictures from the WOW directory on Bart's projection TV. He'd even programmed a scrolling subtitle, "LIVE from the accumulated wisdom of mankind." Much to the amusement of those irritated by Crane's antics, that night the fount of wisdom yielded up a scatter plot of globular cluster velocities, what appeared to be a calibration signal for an interferometry run, and, at the stroke of midnight, an order form for the "Proceedings of the 1993 IAU Vienna Workshop On Relativistic Jets". Sue Hardiman went to the piano and composed a song on the spot. All I remember is,

Crane pursued the aliens, Through the archives on the Net.
But the aliens came and kidnapped Crane,
In their Relativistic Jets.

When I left, they were still adding verses of monotonically decreasing quality. Crane was singing along lustily.

I spent the rest of the week in a happy blur, working on Fantastic Voyager. Every now and then Crane or I would notice something amusing in the WOW directory and send mail pointing it out, but otherwise I didn't pay much attention to the search. By Sunday night, Fantastic Voyager was working pretty well. After tiring of flying lazy orbits around my pancreas, I finished up the documentation describing Fantastic Voyager. Another part of the Week of Rest tradition was that the day the company reopened, papers describing the programmers' projects would appear in everybody's mailbox and the software would be on display in the Demo Lab. After I queued the paper to the mail system, I drove down to the office around midnight to install Fantastic Voyager on the lab machine and make sure it worked there. Also, I wanted to double check that the Quest was set to shut down before 5 A.M. when the Net started charging again. At normal prime time rates, our data transfers for the Quest would've gone through my monthly salary every day, and I wasn't about to get docked a couple of weeks' pay due to a typo.

Crane was in his office, writing up his Week of Rest project, a redesign of an obscure internal algorithm that would speed up one of our products about 30%. He'd actually cleaned the place up, to the extent of even returning most of the books to the bookshelves that covered the walls. We chatted for a couple minutes, then I went down to the Demo Lab and he returned to his documentation. As usual, my demo didn't work the first time--I had to dial into my home machine and copy some files and rebuild. Anyway, it was almost three before I was satisfied it would work that afternoon.

I went back to my office and locked up, then stopped by Crane's on the way out. He'd finished and queued his document, and was idly flipping through the 150 or so pictures in WOW. He said we'd accomplished something worthwhile and suggested we should write a paper about the Quest. Every message we'd found, after all, was the product of intelligence. If we'd broadcast the Net feed to the stars, any civilization with our program would have found ample evidence of intelligence. He thought we might be able to persuade the investigators on the various radiotelescope projects to routinely run our filter over their data, forwarding all the images it found to us for examination.

We talked about how we should phrase such a paper and what journal might publish it, and so on, in the kind of lazy-days bull session exhausted programmers are prone to as they sit on a virtual veranda on the banks of the Mighty Megaflop and watch time and data flow by. "Peep." Another image popped up on the screen: a NASA logo. Crane hit the key to file it away. It was already four, so we decided to sit around and make sure the Quest terminated on schedule at 4:50. Nobody'd be expecting any programmers to show up much before two P.M., so there wasn't any reason not to see it through. Our conversation was describing a chaotic orbit around the usual attractors of evolution, anarchy, programming languages, quantum electronics, and office politics, when we got another "Peep." It was 4:32:19.217 PST.

I can vouch for the time; it's in the execution log. My memory of what happened after the image popped up on Crane's screen is less precise. We glanced at the screen, then we stared at it. The image filled about a quarter of the screen; it was 2039 by 2053 pixels. Four bold black diagonal stripes set off each corner and continued as thin alternating black and white pinstripes along the edges of the image. Had the upper left stripe proclaimed "New!," the layout would've blended indistinguishably into Time or Scientific Enquirer. What was within the border was another matter entirely.

The figures on the right were humanoid, but not human. They were simply the best aliens I'd ever seen: there wasn't a single missing or disproportionate characteristic about them, but they were all wrong to be humans. There were two adults, male and female--the female was holding a little one. On the left was what was clearly a conversion table between binary numbers represented by "-" and "|" and eight digit symbols. The middle of the picture was divided into four boxes, three with a large central dot, curved lines, and nomenclature using the eight digits. The fourth contained what could only be an orrery presentation of a planetary system. At the bottom were four rows of 64 of the digits given at the left of the picture.

Crane wasn't a practical joker and neither was I, but after Lazslo's party all the usual suspects knew what we were up to, and half a dozen of them would enjoy nothing more than sending us on a merry chase. Art rolled his chair over to the screen to check the source. I looked over his shoulder. We looked at each other. Suddenly I knew how Penzias and Wilson must have felt when they concluded that they'd either discovered pigeonshit in their antenna or the birth cry of the Universe. What we were looking at was either the most brilliantly executed gotcha I had ever encountered, or, well...precisely what it appeared to be. Art copied the source identifier, "HGIDB11-23@NIH.GOV" into his notebook, tore out the page, and said "I think you'd better verify this across the Net, and don't use any software we've developed this week; it may have been tricked."

Just to be extra careful, I dialed into one of the commercial compute services the company subscribed to and used standard retrieval and image processing tools in that system's library to extract the image from the location Crane's filter had spotted it, accessing the read-only master published copy of the database across the Net. It took me two hours to be sure. When I returned to Crane's office, the eastern sky was brightening with dawn. It was going to be one of those windy January days when the air is so clear even the distant mountains seem close enough to touch. Yet on this morning, only the weather seemed clear. Crane had also been checking out the Message.

If I'd been more selective in choosing data to scan, we'd never have found it. There could be no doubt; the Message was real, and it was right smack in the middle of where it most definitely didn't belong. We had found it in the NIH Human Genome database.


Every human who'd turned his eyes skyward wondering, "Are we alone?" had been carrying the answer, all along, in every cell of his body. Since then the same Message has been found in every mammalian genome that's been dumped, yet nowhere else in the animal kingdom. That checks with the age of the Message. Three of the boxes in the middle of the Message are orthogonal views of the Local Group galaxies, centered on the nucleus of the Milky Way. Running their motion forward from the positions in the chart to the present day dates the Message at about 250 million years before the present: just about the age of mammals. The orrery indicates they came from the fifth planet of a star with six terrestrial planets and three gas giants. Which star? Nobody has a clue, and what with differential motion over an entire circuit of the Galaxy, any position in the Message would be worthless anyway.

Whoever wove that Message into our DNA meant it to last until we figured out how to read it--it's built inextricably into the protein expression mechanism so any organism with a corrupted copy won't be viable. It's obvious in retrospect. If a visitor wants to leave a message, why not make it a self-reproducing, error-correcting message that any sentient race would stumble upon as soon as they undertook to reverse engineer their own design?

Crane immediately suggested what's become the consensus hypothesis about the number at the bottom of the Message. He thinks it's the product of very large prime numbers and that when we manage to factor it, they'll be the key to decode the second Message. Two problems: first, we either need to learn a whole lot more about factoring, build computers a million times faster than the latest quantum electronic models, or else wait the tens of millions of years it looks like factoring that number will take.

Second, we have yet to find the other Message. Nothing more has been found in the genome, although Crane and many others suspect the factors of the key number may direct us to additional information there. The intensive radio SETI program sparked by the Message has proved barren. So have the planetary explorations of the past 15 years, spawned by the possibility that those who placed the Message may have left a calling card somewhere in the Solar System. More recently, the limitations of robots in seeking the unexpected has triggered the rapid expansion of the human presence into space. If we've found nothing, we've learned much, and perhaps we've taken our first steps in the footprints of those who left the Message in our genes.

Crane believes that when we find the second Message and are able to read it, we'll have proved ourselves ready for the instructions it contains. He thinks it will tell us how to find those who left it, and how to go there. We'll return, after this Message, and I think they'll be proud of us. They were already proud when they put the Message there a Galactic year ago. Proud enough to have signed their work.


References and in-jokes

Calendar for the period of the story

   December 1997               January 1998
 S  M Tu  W Th  F  S        S  M Tu  W Th  F  S
    1  2  3  4  5  6                    1  2  3
 7  8  9 10 11 12 13        4  5  6  7  8  9 10
14 15 16 17 18 19 20       11 12 13 14 15 16 17
21 22 23 24 25 26 27       18 19 20 21 22 23 24
28 29 30 31                25 26 27 28 29 30 31

...Galileo spotted Neptune...

Kowal and Drake confimed the Galileo observation in the early 1980s.

...not in our stars, but on our shelves...

Shakespeare, Julius Caesar, Act I, Scene II, 140/1681

...obtrusive, blatant, ...entice.

Ludwig von Mises, Human Action, Chapter XV, Section 13.

...directory called WOW.

WOW is an in-joke among SETI people. A one-time observation of a signal near the galactic center at the Ohio State University Radio Observatory passed virtually every test for an interstellar beacon and is referred to as the WOW signal. It has never been observed since.

...tens of millions of years...

According to Dorothy Denning, Cryptography and Data Security, Addison-Wesley, 1983, the fastest factoring algorithm runs in Exp[Sqrt[Log[n] Log[Log[n]]]] time, where n is the number to be factored. If we assume a computer that can execute 10^10 primitive factoring operations per second, then the number of years to factor a number n is given by:
            opssec = 10^10
            secyear = 365.2524 24 60 60
            facyears[n_] := (N[factime[n]] Years) / (opssec secyear)
Hence, for our 4 lines of 64 octal digits each, we have the factoring time in years as: facyears[8^(4 64)], or 4.03711e7.