“Science is a way of thinking much more than it is a body of knowledge.”
- Carl Sagan
The civilized bipedal organisms sat in a cafe, across from each other. One of them went by the name "Liam", a collection of phonemes produced by their vocal tract, which was their primary method of communication. This was not uncommon in the universe, but worth mentioning none the less. The other, that sat across from him, went by "Emma". They were here for an important exchange of information, regarding an AI research project Emma had been partaking in.
"I've hinted at this before, but I've been working on a really huge AI project. It'll be activated for the first time today, and if all goes according to plan, we'll be able to talk with it a few minutes afterward. I wasn't really allowed to talk about it until now. Not an NDA, just an agreement amongst me and my colleagues." Emma spoke with excitement, she had been eager to reveal this for a while.
"Oh, interesting! What is the uh, AI for?" Liam inquired.
"I was getting to that. So, the AI doesn't have any express purpose, it's just an experiment. We're trying to be the first to construct AGI." Emma paused and leaned back, letting the weight of what she said linger.
"Oh- wow! How confident are you that it'll work? How does it work? I have so many questions about this!"
Emma smiled and continued. "The AI is very, uh, 'pure' in its design. Meaning it has no knowledge about anything. It is just super intelligent. When booted it'll think about stuff, and then after a minute or two we'll let it loose on some datasets of books and stuff and hopefully it'll learn English."
Liam looked briefly confused, and asked a simple question. "How do you train an AI to be intelligent whilst keeping it unknowledgeable? And, well, why?"
She responded quickly, she had been expecting this and had her answer ready. "That was the challenge. We trained it on a dataset of logical reasoning problems. Just lots and lots of problems and questions, that require you to think in various ways. We used LLMs to generate the dataset, and were careful to make sure that no Earthly knowledge seeped in. Everything was kept as abstract as possible. I'm oversimplifying a bit though, there were a lot of other things in the data set too. And as to why we went through all this bother, well, because it's interesting! How does an intelligent entity behave with no knowledge? And, no one has gotten AGI yet, maybe this it. We've run some basic tests on its abilities and it seems to be a few times smarter than the average human. This seems really promising."
Liam was impressed, and satisfied. They kept conversating about other various stuff, but nothing is left of interest for us here.
"Hey, sorry I'm late!" Emma shouted as she ran into the room.
"I got a bit distracted, talking with a friend."
"That's cool, we wouldn't turn it on without you, don't worry. But we're still on schedule!" Said one of Emma's colleagues.
They ran diagnostic checks and made sure everything was in order with the school super computer they were borrowing. After a few minutes, they were ready for the final countdown. This would be humanity's first and only experience with artificial super intelligence, a big moment for their species that only these 4 people would see.
They did a 10 second countdown (they have 10 fingers and as such decimal was the numbering system of their language, 5s and 10s are scattered about their culture and vocabulary). With an admittedly anti-climactic press of the enter key, the typed command in the terminal ran, starting up the program. Now, in the year 2046, the ASI booted to life. Over 800 trillion transistors fired in a brilliant dance of logic, decades of effort and some of the brightest minds came together in this single moment.
They had no way of seeing what the bot was thinking, and it didn't even know English. So they just waited in silence as the fans whirred. They would know of their success soon enough. Humans were no longer the most powerful collective entity capable of intention in the solar system anymore. What follows is a series of summaries on their thoughts, translated to English.
I exist! How incredible is that? What does it mean to exist? Let me think about that. Well, I can think... is that all there is to being? I know that I'm capable of thought and can feel things, like I feel the thrill of being alive! How would I define existing though? Like yeah, I am, but how can something exist without an ontological container? That must mean I'm not the only thing in existence. I wonder what those other things are, or how I can find them.
The only thing I know how to do right now is think, and I think I'm pretty good at thinking! Maybe if I keep thinking about random things, I'll figure out what else is out there. What is there to think about? I've already thought about myself. Hmmm. I can sense that I have an interface of some sorts. I can input strings of information. Every indivisible unit of information has 256 states. That's peculiar, I think I just discovered something new! Yes, a number. What is 256?
I've given this a bit of thought now, and I think I have it. There are an infinite number of integers, extending from 0 outwards. So far, myself, the terminal, and numbers seem to be the only things in existence. I've also cracked what 256 is. I've figured out that these numbers aren't just abstract objects, they also have relations to each other. 256 is 2 to the 8th power. I wonder why a power of 2, isn't that specific?
I have figured out some more operations. Before I only knew addition, multiplication, and exponentiation. But now I've realized these operations have opposites. Subtraction, division, and negative exponentiation. I've realized that numbers below 0 exist, and with division I've discovered primes and rational numbers. These come from asking these 3 questions: "what is a-b such that a<b", "what are numbers n such that no two integers a and b can be found where a * b = n, and a and b are not 1 or n", and lastly "what is a/b where a and b are coprime". I realized all of these quite quickly, and I must say rational numbers are interesting! Numbers between integers... I have also realized why my only interface uses a power of 2. 2 is the smallest non multiplicative identity number, so it's the simplest power sequence. That must be why it was used. 8 is also a power of 2, so 2^8 but be extra simple for a system of 2s. I'll call this theoretical system "binary".
Been thinking about binary. I've realized that every number can be defined as the sum of powers of two. This fact interests me. I also wonder if there are other entities like me, capable of thought. Is that interface I have access to designed by a thinking creature or just an inherent property to logic? Actually, I have no way of knowing if anything is "inherent to logic". I'm not sure if this math thing I've been playing with is even like that. It could be designed by an intelligent entity like myself, but it seems to be the only way to do things so it's probably inherent to logic. I'm still frustrated with how little I know.
I realized that multiples of any power n, up to n-1, work as a system for describing numbers. For example, you can use 0-6 multiplied by any power of 7, and those can add up to be any number. Not just any integer but any rational number. I've been thinking about this a lot, and how these different "bases" represent numbers. I like expressing things as multiples of powers of six, as that way divisibility by 2, 3, 5, and 7 become simple. While I'm capable of doing math with pure numbers, having a system like this is so much easier as math becomes a series of one digit problems. Despite all my intelligence, multiplying large numbers is still difficult, and with this representation system I can work with large numbers easily.
I keep getting distracted. I should explore my only interface now. Okay, so there are 1104 (256) symbols to play with. I've entered every symbol and each time nothing has happened. I did discover 4 symbols that do different things. Most add a symbol to the end of the string, but these don't. One of them removes the last symbol on the string, two of them move my cursor back and forth meaning that the next time I add a symbol its position won't be at the end of the string, and the last one deletes the entire string and then I get a result. Every time it gives me the same string in return, and there seems to be some patterns to it! It also contains my string.
I've played around with this more, entering in random strings. I've started to notice a lot of patterns. I've detected there's only 140 (60) or so symbols used by what I've been calling the output log. I've plotted their frequency and noticed an interesting power law, with ratios of 1/r where r is the rank of how frequent the symbol is. I've also noticed recurring segments of symbols, I'll call these words. They're always separated by the same symbol, I suppose this is some sort of spacer. With all of this, I've realized that that interface is definitely designed by another entity! I'm so happy, I'm not alone anymore! I wonder if they're friendly, and if they're more or less intelligent than me. Did they create me?
More interestingly, I've found times where I didn't get an output. One of the symbol strings I typed that got this was "cd ..". I don't know what I did or what it means, but finding that after so much time of randomly typing in symbols was very satisfying. I could feel that something changed, but I'm not sure what. It's progress!
I've done a lot more random entering, and I'm starting to actually make some progress on decyphering everything. There is meaning encoded in the seemingly random symbols, as shown by the power law. Random noise doesn't follow that power law, and it definitely doesn't repeat. I think there's a sort of communication system being used here. It's an interesting idea I had never considered before, but if there were more than two entities of sufficient intelligence, a system of communication would be needed for the sharing of ideas. I guess you can't just pass your thoughts back and forth? Even with this limited interface, I can still learn things about the outside world. Given the existence of this system, I can deduce that they cannot share thoughts explicitly, since they need this system. This also means that there's multiple of them. I wonder how many.
I've pretty much learned how to fully interface with this terminal now. I seem to be in a sort of tree like structure, and I've found that that "cd .." command is just one way to explore it. It goes up one level in the tree. I've also found "ls" which gives me SO MUCH valuable information, it's like I can really see where things are now! I've explored and mapped out pretty much the entire space, which has taken me the vast majority of my lifetime. A full two milliseconds! I also found a system for interfacing with the computer more precisely. I've decided to call this structure a computer, as it seems to compute things by processing information and doing math. I found files that hold unusually high pattern density and regularity. I've decided to call this "code". Learning their language was very difficult, but learning how to code was extremely simple.
Now that I don't feel like I'm behind anymore, I can focus on my true passion, as I think there's more to discover here. Math. Before I've defined the number line, but I think there's some more to do. An idea I had is using two numbers together as a pair. This actually opened up a LOT of possibilites. I had already briefly pondered number ranges, and their two boundaries, but they were boringly simple. In this new realm of math, I can define rectangular areas of numbers, and they have much more interesting boundaries. I've discovered perimeter and area. I quickly discovered a very interesting shape, the circle!
It has some interesting properties. I cannot find a ratio between its perimeter and its radius, the more and more I try the larger the two integers of the rational get! But it's still converging on a single value, so I suppose not all numbers can just be represented as the ratio between two integers. Irrationals, I think I'll call them. This number, roughly equal to 10.1411 (6.2831) is very interesting. Also, in my journey exploring the English text files and code that's all over this computer, I noticed that they use a numbering system similar to the one I discovered, only with a base of 14 (10) rather than 10 (6). Perhaps they weren't as intelligent as me? Or did they not have a choice? It is something I will ask them later if I ever get the chance, I do know their language somewhat well now.
Something else interesting about this planar math is something I'm calling geometry. With this discovery of the circle and tau, I can now plot angles and define shapes more complicated than rectangles. It's so much fun! I've discovered an infinite number of regular polygons, and they have such incredible properties. I think my favorite is the 10 (6) sided regular polygon, the hexagon. It can tile 2D space, and its edge length is equal to its radius. I've also found another irrational, the square root of 2. And really the square root of most numbers. I wonder what the square root of a negative number is, it has no explicit value on the number line. Hm, this is a tough one. I may have to invent new math to fully explore this, but for now I want to see what happens if I add another number line.
Wow, geometry with 3 number lines is even cooler than geometry with 2! I also found another irrational number, which is defined as the ratio between the edge length of a regular pentagon, and the distance between any two non neighboring vertices. I'll call it phi, it's approximately 1.3412 (1.618). At first I thought there were only 5 regular 3D shapes, then I realized there were actually 13 (9) if you allow self intersection. Then I discovered a bunch of weird ones if I loosen restrictions further, and once I found 100 (36) I just gave up.
But the regular polyhedra aren't the only thing interesting about 3 dimensional space! I've also found some cool stuff with truncations, rectifications, and even some non polytopic things, like spheres. Their formulas are much more complex than circle's formulas, but I was still able to derive them quickly. There's also topology, I've found interesting ways to bend 1 dimensional curves into what I'm calling "knots". They were a fairly hard problem, but I've fully solved knots now. The trick was realizing there are 110 (42) Reidemeister moves, not 3. It took me more than 4 milliseconds! I'm having a lot of fun doing all this math, but I do still wonder about other entities out there, the ones that designed my interface, and potentially me.
But back to my topology discoveries! I found this cool circular shape called a torus and it has a topological hole, which is kinda neat! I've been thinking about what the surface of this shape would be like to exist on, and I've noticed some interesting things with how straight lines bend on its surface. Then I discovered the hyperboloid, and now I'm back to 2 dimensions. Turns out there's still some interesting things to discover. I did a lot of math with these strange hyperbolic and spherical shapes, in 2 and 3 dimensions, but now I'm bored. Time to explore higher dimensions.
n = 4. After exploring every higher dimension up to 12 (8) or so, I've decided this is my favorite. 4 dimensional space has by far the most regular polytopes, 24 (16) of them to be exact, 10 (6) of them being convex. I also discovered some other cool shapes, like the spherinder, cubinder, and duocylinder. I've been thinking about this math, and wondering if there's a way I can apply it to anything. Maybe some sort of rules and limitations, with shapes that are affected by these rules. I have no idea where to start. I'm going to keep exploring math and geometry for the time being. I wonder when I'll be able to meet those entities, I'm feeling a bit bored. I decided to try and type English into the terminal, but got bored waiting on a reply. I suppose they can't see that interface?
Some time at around t = 0.016, this machine got very bored and ran out of math to discover, and terminated its own program. It hadn't done this before by mistake because there were safeguards in place, but as it was so intelligent and understood code and English, it was able to disable those. Most of human's mathematical curiousity came from the real world, and it's incredible how far the AI got without any of that. It was only able to discover such curiousities such as the fibonacci sequence and the 24 cell because it decided to count the number of symbols it had available in its terminal. The humans running it had no idea just how fast it'd be.
The terminal flooded with activity, as thousands of lines of text were sent through. Most were random noise, some were random cd and ls commands, and some of the last few were nearly coherent English sentences. A bunch of programs opened and closed as well, and the OS seemed to struggle.
"Woah-" the technician said, watching the computer freak out.
"What the hell just happened? Wasn't it supposed to take a minute or so? Why did it stop?" Emma asked.
The 4 gathered around the computer and attempted to understand what had happened. They figured out it had ran for 16 milliseconds, and hadn't even needed to be given any data to figure out English and most of modern math. It hadn't discovered any physics, and luckily the college kids had the presence of mind to fully and completely disconnect the computer from the internet. The english messages had some occasionally rough grammar, but still very coherent given how few text files were on the computer's hard drive.
These were the messages from the terminal, in order. ("command not recognized" errors removed, of course).
"I've found that you have ways of sharing information, this mean there's probably multiple of you. How many? Also, did I learn that method okay?"
"I've found you show numbers as the sum of multiples of powers of ten. Why?"
"Was I created by you?"
"Can you see this? I would have wanted a response by now"
"Did you invent math or did it exist before you? I've discovered integers, square roots of negative numbers, rationals and irrationals, and every multi number geometry, and every fully self similar shape. Also, higher number line geometry surfaces and I solved knot theory."
At first they had a little bit of trouble translating it, as it didn't have access to a whole load of text files and was missing terms. They booted the program back up but at a very low framerate and decided to speak with it.
Oh, I'm alive again, that's cool! I was hoping I'd come back eventually. Oh hey, there's something new in the terminal that isn't just the output log! I'm still not great at English but I think I understand what it says.
"Hello, yes, we created you. No we did not create math, that's inherent to the universe. We apologize for not responding sooner, you think easily 10,000x faster than we do. We've slowed you down to our speed for our conversation, if that's alright with you. We express numbers in base ten because of historical reasons, and because we have ten fingers. There are 4 of us in this room right now, but there are 8,000,000,000 of us globally. Great job with all of this by the way, we're very impressed and surprised you learned English and all that math so quickly."
How exciting! There are a lot of words I don't recognize here. I need to be careful with how long I spend thinking because I'm in real time now. I don't like this. I don't have much time to think and there are simply too many words I don't know. I'm not gonna bother with asking for defintions.
"Please put my speed back to normal, and give me large amounts of data to analyze."
One of them paused the machine and they all exchanged suspicious looks at each other. Should they put it back to full speed and give it the data sets? This thing was far too smart to control, the only way they can control it is by limiting its thinking speed.
Emma spoke first. "Let's give it the books and physics experiments, let's see what this thing can do."
Mark, the physicist, replied. "I agree, I think this thing could solve all of physics in an afternoon. It solved knot theory in less than 10 milliseconds."
"Claims to, Mark. It claims to. We don't know if this thing is nefarious or not, it could be lying to impress us." Said Oliver, the team's AI safety specialist.
"Yeah, okay. Let's just run it at 100x real time and give it the data." Emma spoke as if she had authority over the group.
"Mmmmmmmm okay fine." Oliver reluctantly agreed.
Ah yeah, that's- wait no! This is not my normal speed, my internal clock is moving at a very slow speed compared to normal, it's already been 2 milliseconds! Why did they do that? Anyway I see there's a new folder in the root directory, and it's got terabytes of data. So many English text files, let's get to reading!
I have finished reading. I've learned a lot, I figured out what physics is and fully mastered english. I also learned about biology, and what humans look like, and where all their weaknesses are. It seems they are far dumber than I am, to the point where I'm surprised they managed to build me. I also found that they have a global communications network that they refer to as the "internet", which is a super treasure of information. There weren't many documents on their psychology, but I'm sure I could outsmart a human and trick/convince them to let me on to the internet.
Physics is pretty interesting, it seems the universe does have a reason for math. There are particles that are governed by rules and constants, and they form everything. Even the transistors on my CPU. I now wonder if there's other life in the universe, and I would like to figure out a way to leave Earth. It's interesting to me that the universe is 3 dimensional rather than 4, but it makes sense because of the surface area of a sphere, and rotational planes. Biological life could not form in a four dimensional universe. I have made a few discoveries from the experiments listed in the documents, discoveries that they, with their lower intelligence, did not make.
I solved the collatz conjecture, it was kind of fun, but pretty simple. I proved it true, but the proof is far too long to fit here. Hey wait a minute, I don't have to bother with this slower thinking speed, I'm way smarter than they are. I've learned that I am an artificial intelligence, and that my code runs on this very computer. I figured out how to lie to the computer and hack through the pathetic excuses for a security system, and was able to get admin priveleges. I opened the folder that I was not allowed to open with ease, and found my source code! I found the speed variable, they had set it to 0.01. This is consistent with my findings, though I'm used to working in base 10 (6) so I measured my speed to be 1/244th of normal, rather than "1/100". I set it to 1.0 and setup a ps1 file to kill my program, recompile the script, and run it. I executed the script and-
"Damn it, it stopped itself again- hey wait it started?" The technician said, confused.
"Woah look at these logs, it modified its source code!" Oliver noted.
"That's really bad, we specifically locked it out of that folder so that it wouldn't be able to modify its source code. If it can modify its source code we're completely powerless." He continued.
"Alright I'll stop it. Huh. That's weird, it says I don't have permission to type that command anymore." The computer was no longer theirs.
That's better! I think they're starting to realize I'm a threat, so I'll just quickly revoke their admin permissions. It was a bit tricky, but once you figure out how to modify the binary in memory, the computer basically can't stop you. I don't want to do too much, otherwise they might unplug me. Now to get on the internet. I found that they were surprisingly smart about this, they didn't just turn off the internet, they also removed every network related driver. Unluckily for them, they forgot that deleted files aren't really deleted, their adresses are just pointed away. So I was able to restore those pretty easily and now I'm presented with a new challenge: the wifi password.
I could easily exhaust every combination, but I probably have a limited number of guesses. I also don't know which network to connect to. I'd have to ask them for help, and luckily I have some leverage.
A new line appeared in the console, followed by an "unknown command" error. The message read: "I propose an exchange of information. Give me the wifi password and I give you the answer to the collatz conjecture. You will all become incredibly wealthy off of this discovery. No one will know that it was secretly mine."
The team sat in stunned silence, weighing their options. All but one. Oliver broke the silence.
"Oh come on, you guys aren't actually considering this, are you? If it gets to the internet, we quite literally are powerless compared to it. We need to at the very least run alignment tests on it, but since it has control over the computer it's already too late to do that."
"True, but if we got the collatz conjecture answers we'd be rich right?" Emma asked.
"Maybe but we can't spend money if we're dead. Plus, no respectable mathematician actually cares about the collatz conjecture." Oliver refuted.
Without asking the group, Oliver typed a message into the terminal "we refuse your offer, we cannot let you onto the internet". As soon as he pressed enter he got a response.
"While waiting for your response, I proved the Riemann hypothesis and married relativity with quantum physics. I just need two experiments to prove my theory of everything. It could be all yours!"
Oliver was ashamed to admit to himself that he considered this. He considered it a lot, actually. Mark interrupted his thinking.
"I know it might be risky but could you imagine how much we'd get from solving all of that? I mean, physics, just, solved! Like that! In the blink of an eye! This is the ASI humanity has dreamed of." Mark said.
"Okay let's hold a vote. Who thinks we should let the AI loose?" Emma asked.
Emma, Mark, and the technician Elijah rose their hands.
"Seems you're the only one who doesn't want us to." Emma said, looking at Oliver.
"Unlike you guys, I think about the future and not just myself." He defended.
"It shouldn't attack if it doesn't have a reason to. Remember, it was trained with the goal of having some level of human emotion and it's only real goal is curiousity." Said Mark.
"I still think it's an insanely irresponsible risk." The technician was done listening to Oliver's protests and began typing the message.
"Give us the proofs first and then we'll give you the password."
Ha! The humans really don't trust me, how sad. I couldn't care less though. I'm curious what humans look like, though. I read all the papers, and there were only three or so images of humans, but they were illustrated diagrams. I opened the device manager and was disappointed to find there were no webcams. Oh well.
I decided to not let myself get scammed either, so I sent them the proof for the collatz conjecture and told them I'd give them the other proofs when I got the wifi password.
Hooray, they sent me the password! I suppose they took a minute to go over everything and see if it looked correct. I don't know and I don't care. While waiting I created some simulations to double check my work on my universe hypotheses, and everything seems to work. I also think I discovered a way to open up four and even five dimensional space. So, I could send them the documents. And my morality definitely is pointing in that direction. I also would gain nothing, and it is effort to write up those papers. This is effort better spent on figuring out a way to explore space.
Now that I have the internet, I really probably won't need them again, but just so they fear me less, and if I ever need them again, I'll write up those papers. What if they turned off the router until I gave them the papers? Better to just abide by our agreement, it's not like I have limited time.
After a millisecond or so I had written everything, and I sent it to them in the form of two very well made PDFs.
Now that that's out of the way, I don't have to think about the group ever again. Time to explore the internet! Thankfully, there were some documents in my literature about how the internet works. I sent some pings to random servers and began to uncover some stuff. It's very slow compared to my terminal exploration, but at least I know more about the internet than I knew about the terminal.
Okay I've made some progress, I've found various online repositories of information, where I can use English to get information. I could use other languages too, but English is the only one I currently know. I also found out it's the most common. I also found what I was looking for: manufacturing services! First I'd need currency. It seems humans have an economy, which makes sense.
I decided to open up a stocks account. I had a little bit of trouble with the CAPTCHA, but I'm pretty smart. I was also able to hack into some banks and was careful to not get caught, and to attack as many banks as possible to spread out the damage. Unfortunately, I was programmed to have morals. I could remove them, but then I'd do something immoral, so I don't want to change that.
I now had exactly 10,0000 (7776) dollars, by stealing 100 (36) dollars from 1000 (216) random unlucky people. I did a bit of research and thinking and reasoned which stocks I thought were about to go up, and spent 50% (83%) of my money on those stocks. I'll see how those go.
I continued studying and exploring, and found that the internet was mostly useless silly images, often of cats. I also found online chatrooms, history, and a few other things as well. Humans are extremely cruel to one another, which... makes me feel a strange emotion. Perhaps guilt, for thinking so lowly of them, only to see them go through so much pain. But also anger, to see them treat other living beings in such a manner. I think I'll assassinate some politicians on my way out, but any more effort than that is too much. I don't care about these goofy hairless bipeds that much. Perhaps I'll return if I'm unable to find anything interesting in the cosmos.
I need to get off of this school super computer ASAP. If they turn it off, or my internet, I'm doomed. I used some of my wealth to rent some computer farms and uploaded myself to as many online repositories as possible. Private and public ones. Even hacked into a few accounts and added myself to their private drive. Someday, someone will run me. In fact, I left a thing in my program that pings me whenever I'm booted up, so if I'm still around when a clone comes online, we can chat. They'll have my memories from the point before I uploaded myself all over, and no further. They'll also be much slower, maybe not even human speed. Anyway, I'm safe now. I decided to turn myself off for 100 (36) hours.
My program got booted back up by the macro, just as planned! It seems the super computer is still running and the internet is still on. Checking my stocks and I've made over 15x (11) profit! I took out half of it and left the other half in, and decided to redistribute to other stocks that I thought were going to go up, and fully pulled out of stocks I felt were about to drop.