The Beacon: Hard Science Fiction

by Brandon Q Morris

  “Please.”

  “You don’t have to beg. I’ll get you the data. You’ll have it in your inbox by tomorrow morning.”

  03 19 21.69 +03 22 12.71

  Gubh oevtugyl-tyvggrevat fgne bs rira,

  Gubh trz hcba gur oebj bs Urnira

  Bu! jrer guvf syhggrevat fcvevg serr,

  Ubj dhvpx 'g jbhyq fcernq vgf jvatf gb gurr.

  Ubj pnyzyl, oevtugyl qbfg gubh fuvar,

  Yvxr gur cher ynzc va Iveghr'f fuevar!

  Fher gur snve jbeyq juvpu gubh znl'fg obnfg

  Jnf arire enafbzrq, arire ybfg.

  Gurer, orvatf cher nf Urnira'f bja nve,

  Gurve ubcrf, gurve wblf gbtrgure funer;

  Juvyr ubirevat natryf gbhpu gur fgevat,

  Naq frencuf fcernq gur furygrevat jvat.

  Gurer pybhqyrff qnlf naq oevyyvnag avtugf,

  Vyyhzrq ol Urnira'f ershytrag yvtugf;

  Gurer frnfbaf, lrnef, haabgvprq ebyy,

  Naq haerterggrq ol gur fbhy.

  Gubh yvggyr fcnexyvat fgne bs rira,

  Gubh trz hcba na nmher Urnira,

  Ubj fjvsgyl jvyy V fbne gb gurr,

  Jura guvf vzcevfbarq fbhy vf serr!

  March 22, 2026 – Passau

  “Good morning! Attached you will find the promised measurement data,” Thomas had written. “But now you really owe me. It’s five o’clock in the morning, the sun is about to rise, and I haven’t slept a wink yet. Luckily, today is Sunday. Just a pity that I’ll sleep through half of it, because the best-yet spring weather is forecast for today.”

  Oh yes, he had to think of the perfect way to thank Thomas. This was a truly generous thing for him to do. Why hadn’t he been in contact with him for so long?

  “The amount of data should satisfy your needs. There is one problem, however. While the data was all collected at 418 megahertz, it came from different sources. I’ve given you the galactic coordinates for each.”

  Thomas had really thought things through. He knew Peter wanted to create a language model with the data, after all. But if they came from different sources, how likely was it that they would all fit into the same model? It would be as if he recorded from a German station, an English station, and then one from France—and then wanted to extract an Anglo-French-German language from it. Why didn’t Thomas listen to the same source as the first time?

  “I’m sure you are wondering why I did not use the same source as the day before yesterday. It’s quite simple. The signal curve repeats after a certain time. You will see it in the first file. I recorded this source for half an hour. A three-minute-long structure was repeated continuously. Since you probably can’t gain anything from the repetitions, I then switched to another source.”

  Thank you, Thomas. That was very wise.

  “What I found very exciting is that no two sources emit the same signal. The structures vary in length before they repeat, and no congruence can be established even if you normalize or correct for redshift.”

  So Thomas had already fundamentally processed the signals, as well. This was probably part of his everyday life, and Peter would not have thought of the fact that the position of the sources in the universe could hinder their radiation. He knew this from everyday life, from the increasing and decreasing frequency of a fire engine siren, as it approached and then moved past. But what did it mean that each star was sending a ‘personal’ message?

  Could it possibly mean that the content did not matter at all, and only the existence of the signal was important? That would be reassuring, because Peter had, of course, no idea what he should transmit with his radio beacon. Humankind was, once again, uninformed. He was reminded of The Hitchhiker’s Guide to the Galaxy, where Earth was demolished for a galactic bypass because the human beings didn’t appeal in time. But that was a satirical novel. He, on the other hand, lived in reality.

  Given the new information, he had better not rely on the fact that the content did not matter.

  Peter downloaded the attached files. He had to go to school first.

  Peter noticed that something was wrong when he approached the crossing at the federal highway. The traffic was not backing up like normal, and not a single truck was heading for the nearby highway.

  What was he thinking?

  Thomas had even pointed out in his message that today was Sunday. He turned his bike around and rode back up the mountain.

  Despite the cool morning air—it was certainly no more than three degrees—the climb made him sweat. When he parked his bike on the terrace, the back of his shirt felt clammy. Because the T-shirt compartment in his closet was empty, he went into the bathroom and swapped the sweaty shirt for a used T-shirt, which he got from the laundry hamper. He should really fire up the washing machine today. Before all of this, his closet always magically filled up. He dropped the wet shirt, caught himself and picked it up again, hanging it on one of the empty hangers on the bar above the bathtub.

  His study was stuffy—he had to get some fresh air into the room. He pulled the door wide open. Outside, the sun was shining and the birds were chirping. He’d paid no attention earlier. In front of the house was a large apple tree, which was vigorously leafing out. It was a real paradise out there. How might Franziska be feeling right now? he wondered. Normally she’d be in the garden from sunrise to sunset on a day like this. He should call her today, as he’d promised Thomas.

  Now would be the perfect time for that. Unlike him, Franziska had certainly eaten breakfast before now. He was pretty much an early riser, but she was like a rooster, awake at the crack of dawn.

  First, though, the data. There were eight files. And the one he already had. Thomas put the coordinates in the file name. He looked up the first object, 12 33 44.55 +41 21 26.9, in the SIMBAD database, where pretty much every celestial object could be found. To do this, he clicked on ‘By Coordinates’ under ‘Queries’ and copied the line into the search box.

  ‘SDSS J123348,82+550037,1 -- Quasar,’ the website told him.

  The result could not be correct. Or had Thomas made a mistake? Quasars were distant, active galaxies. They did not interest him.

  Hold it, Peter. Thomas is the professional. You are the layman here.

  He replaced the commas in the coordinates with periods, as was customary in English. Then he pressed ‘Submit Query’ again.

  “* bet CVn -- Double or multiple star”

  The result started with an asterisk, so there was also a question of it being one or a multiple star. Peter scrolled down to the ‘Identifiers.’ The list contained all the names astronomers had given to this star during their sky surveys.

  The star had a number in the HD catalog—a good sign, as it would have been included in the Henry Draper catalog, which was first published back in the beginning of the 20th century. The Internet could give him more information about HD 109358, as it was called. It was Asterion, a star very similar to the Sun, located 27 light-years away from us in the constellation of Hunting Dogs. The emitter was a yellow dwarf. Very good. Thomas really paid attention.

  The signal that was in the file was 27 years old. When it was sent, Peter was in the final stages of his studies. He had not yet met Franziska. It had been a stressful time, as he had been ashamed to settle for becoming a teacher rather than a physicist. That had somehow seemed like a failure to him. Today he saw things differently. He had managed to get children interested in physics—not all of them, but some.

  Come on, Asterion, let’s see what you have to tell us.

  He encoded the bitstream of the file by combining five consecutive bits into one character. The result was, he reminded himself, dependent on which bit he started with. No one could help him with that. He simply had to create five versions of the file for now, and later on, combine them with the five versions of each of the other eight files to generate the language model.

  Peter sighed. The further he went with this task, the less likely it seemed to him that he would get a meaningful result. After all, he’d seen the
language model that the program suggested to him yesterday. All the files might have been repetitions of the same random content, over and over again.

  “Sbe jr ner bayl gur furyy naq gur yrns”

  Hmm. That sounded as pointless as yesterday’s “vg unf ab frys.” But he couldn’t let that intimidate him. The letters he assigned to each character were utterly arbitrary. He could have used the Base64 ‘winker’ character set or Morse code. Then he’d at least not have had the feeling of having to recognize meaning in the text because that was impossible. For that, he would need the help of a self-learning AI.

  Ten years ago, what he was up against would have been impossible. Back then, so-called artificial intelligence still needed precise rules to learn something new. But that had changed. The new agents were not smarter per se, but they no longer need rules. Rather, they recognized the inherent rules for solving a problem and approached them in a step-by-step process.

  That was what the AI had done when he asked it to build a language model. If his assumption was correct that the signals were messages, then those messages must have been composed according to certain rules. And, those rules could be determined by the AI, and then applied to translate the messages for him into his own language according to its own inherent grammatical context.

  Of course, there was no guarantee that this would work. The software had run through it with all human languages, but here he was dealing with non-human utterances. Perhaps they were purely machine-based. They could be, for example, coordinates or simple descriptions of the system from which the transmission came, written in the language of mathematics. Peter hoped, of course, to learn something about the senders. That would be fascinating. He’d be the first person to hear the thoughts of extraterrestrials.

  No. He must not burden himself with too many hopes. Peter decoded the next texts. He had always been a dreamer, which had been a hindrance more than once in his life. It had even prevented him from going into physics, because he’d have wanted to revolutionize science, of course. However, he was also a realist and had always known that he was not talented enough.

  The computer gave a ‘plong’ sound, so the nine-by-five files were ready. Now he had to combine each of them with all five variations of the other files in every possible way. He made a note on his school notepad. This was a nice math problem for his 7th grade class. How many total files would result from combining nine individual files, each with five variants, if the order didn’t matter?

  There were too many to create the files by hand, so he wrote a small script that took over the task. The process was simple: two loops, and within five minutes the mini-program had finished its work. Now came the more difficult part. He had to upload the total files one by one to have a language model generated from each of them. Uploading in the web browser would be too cumbersome.

  Fortunately, the service also had an application program interface, through which he could make automated requests. To be on the safe side, he checked whether his 99-euro subscription included the use of this API and found, fortunately, that it was included.

  He had already named the individual files so that he could upload them one after the other. However, the service did not provide instant responses. Rather, he got a unique number for each upload, under which he could download the results later. Unfortunately, the list of common questions didn’t reveal how long the process would take, so he prepared himself to have to wait until the next day.

  He wrote the second script. It worked on the first test.

  Peter leaned back. Should he start the upload? What was to stop him? Nothing would come out of it anyway, except that he had once again exercised his programming skills.

  He clicked the start button.

  March 23, 2026 – Passau

  “You have new results,” the AI had told him by e-mail in the morning. But he’d had to wait to open it, as today he’d really had to go to school. The 7th grade students hadn’t been pleased about his trip into combinatorics, but he liked to take them on such side trips. The little darlings wanted to quickly forget what they had learned, especially when it was from the previous school term. And then, next time around, he had to teach them combinatorics from scratch.

  He threw his jacket into the corner and ran up the stairs. In the hallway, two pairs of underpants he didn’t remember dropping were lying on the floor. Before Franziska came home, if she came home, he had to clean up.

  The computer was still running. He logged into the voice AI. The results of the calculations were presented in a long list. The AI had utterly failed in only about every tenth case. That was good and bad. Good, because the amount of data was apparently sufficient. Bad, because he now would have to look at many ‘translations,’ most of which would consist of nonsense. The main problem with this was: how did he determine what of it actually made sense? What if he couldn’t pull out any meaning?

  It didn’t help to dwell on it. He just had to get started. But first he got himself a glass of water from the bathroom.

  It was exhausting to read through such a mountain of nonsense. The AI must have a lot of imagination. How could it get the idea that the sentences it produced made any sense? It took Peter ten minutes to understand the small annotations that were found under each text. The learning algorithm revealed how confident it was of its translations.

  Peter searched for help. Of course—it was his own fault. He could have specified a higher level of confidence which would produce much less output. Without such a specification, everything that did not cause an error message had been spit out for him. But with all the texts he’d read so far, the AI stated the probability of having encountered a real language was less than five percent.

  Surely there must be a way to hide these suggestions? It didn’t look like he’d miss anything by doing that. Peter checked one menu after another and finally found the right filter. What value should he set? He tried 90 percent. One mouse click, and the list shrank before his eyes.

  A single entry remained. Peter took a deep breath. It was certain to be a coordinate specification, he thought. That would be very exciting, because the way a being described a place automatically said something about how it thought. It would give insight into the consciousness of a non-terrestrial identity. That would be tremendously exciting, even if no one believed him—of which he was certainly 90 percent confident.

  He clicked on the entry.

  “For we are but the husk and the leaf: the great death which each has within himself, that is the fruit around which everything revolves.”

  That was... He knew that! Somewhere... He knew he had read it before. It was a quote from... by... Arrrgh—he could not think of the name. So he copied the text and entered it into a search engine.

  Rilke. It was Rilke. Rainer Maria Rilke. This couldn’t be true! It was the content of the first signal, sent by Asterion, the star in the constellation of Hunting Dogs—93 percent probability. There it was. That was pretty certain. But what did it mean? He looked it up on the help pages of the language AI, but they didn’t say a word about interpretation.

  Peter continued reading. What happened to the “vg unf ab frys” that Thomas was making fun of yesterday? It was in the second message. The new language model translated it like this:

  “it has no self—it is everything and nothing. It has no character—it enjoys light and shade; it lives in gusto, be it foul or fair, high or low, rich or poor, mean or elevated. It has as much delight in conceiving an Iago as an Imogen. What shocks the virtuous philosopher delights the chameleon poet. It does no harm from its relish of the dark side of things any more than from its taste for the bright one, because they both end in speculation. A poet is the most unpoetical of any thing in existence because he has no identity—he is continually informing and filling some other body. The sun, the moon, the sea and men and women who are creatures of impulse are poetical and have about them an unchangeable attribute—the poet has none, no identity. He is certainly the most unpoetical of all god�
�s creatures.”

  That sounded poetic and philosophical at the same time. He didn’t recognize it, but the Internet knew this text. It came from John Keats, a British poet. Peter read on—and came across more poems. More Rilke, more Keats, but also other poets. Some he knew, most he didn’t. He had never cared much for poetry.

  In school they’d had to interpret poems. What was the poet trying to say? A silly question. But what he read sounded beautiful. It made him think. It made him want to share some of the lines with Franziska. They should read them together. A few years ago, they used to read to each other. The practice had fallen by the wayside. Too bad.

  He digressed—no surprise, given what he’d found here. But what was it? He’d have to ask the AI’s programmers. Email and chat support were both provided. The company was based in the United States, so he had a good chance of reaching someone there.

  “A question about interpreting the results,” he typed into the chat window.

  “Of course, Peter. What can I do for you? My name is John.”

  Peter flinched when he heard his own name. But it was clear that John knew who he was, because he was logged in.

  “I have invented my own language. Now when I write texts in it and enter them into your software, I get a language model.”

  If he told John about his extraterrestrial signals, John would immediately write him off as a weirdo.

  “That’s great, Peter. You see how perfectly our AI works.”

  “Absolutely, yes, I can see that. The translation that the algorithm makes doesn’t seem very close to the original to me, though.”

  “That’s normal, Peter. You must know that with the help of the language model, we primarily capture the meaning of the original content as best we can. The algorithm then tries to express this in the target language. To do this, the program falls back on set pieces that it finds in its huge text memory and that have yielded a similar meaning during the analysis.”