Data Mining
"The Freud patch is beginning to do the trick, we believe. It took some adjusting, but overall they are beginning to open up again. All indications suggest that the situation will only continue to improve."
"Nothing conclusive, yet. But we are fairly confident that this is an extension of the Babel glitch, and the Greater Civ crash from last year."
"Well, not exactly. It's less of a consciousness trying to survive our efforts to patch it out, and more....well, it's more a result of the system itself. It's only doing what we told it to."
"Of course you don't. This isn't something that we could have predicted. Here, let me put this all in perspective. You know most of this, but let me give some context, ok?"
_________________________________________________________________________________
Project Narcissus began just over a decade ago with one particularly robust mind asked a simple enough question: What if we could build a simulation of our own universe but somehow speed the development of things up dramatically? Could we potentially see the future of our own universe play out in front of us?
Captivated by the potential, this mind brought in other minds, and together they began developing the simulation. When word of the project got out, though, some very predictable issues began to arise. Business giants, military leaders, and religious institutions began to harass the team of minds for access to their findings when they finish. At first they tried to explain the problem with this kind of thinking.
They eventually gave up on explanations, though, and simply responded to each request with a polite but firm, "No."
There was no fanfare the day they turned it on. No press hearings, long speeches, not even champagne. One of them simply muttered to themselves, "Let there be light.", and pressed the power button.
Things went well at first. The development of the first elements and even many of the astral bodies took up next to no processing power, and so went by quickly. As things got more complex, though, the machine began to slow, if only by a little. The Cambrian explosion came and went, the dinosaurs rose and fell, and eventually some relatively small and unimpressive apes came down from the trees to begin building civilization.
When they began speaking to each other, though, the machine began to lag tremendously. The minds weren't sure why. The words should be nothing more than text data, logged in a file as spoken. Sure, the file was getting larger and larger, but overall it was nothing compared to the calculations running to keep the cosmic bodies these apes would never even see in the right place at the right time. So why were the spoken data points having such an impact?
The minds never really figured it out, but they did find a fix. They broke the files into multiple files, creating "languages" that corresponded to each file. They called this the Babel glitch, kept thorough records, but ultimately moved on without too much fuss.
As the civilizations grew and became more complex, however, the problems began again. The apes were now using the first cars, telephones, airplanes, and advancing every day. They could speak to others who were very far away, and they could get to them much faster. The minds were asleep when this started, and by the time they woke there was a war involving most of the planet only coming to a close. The minds, obviously concerned, wondered if another patch would be necessary. Right as they agreed that the issue seemed to have resolved itself, a much worse conflict began. The issue was patched again, but not before the simulation nearly destroyed itself.
They decided to slow the simulation down some, and keep a more watchful eye on it. Generally speaking, though, things went smoothly (or, as smoothly as they could after an event like that) from thereon out. Until this week.
_________________________________________________________________________________
"I've been running some analyses on the spoken data points, and there's something strange in the files. There are the 6,500 or so files that we created to divide the data logs into smaller chunks, of course. But at the very bottom of that list I found another folder that we didn't put there labelled 'Subtext'. See, for each word logged in the spoken data points there are thousands of associated data points in the subtext files. Literally every word has a sister log that is massive. And it isn't just language. There are emotions, sounds, images, memories, entire concepts and world views contained in just one of these words in some cases. What's worse, the data expansion isn't limited to the words themselves. Each simulated person has their own 'package' attached to almost every word they use, and while most are similar for identical words, none are exactly the same. If we want pure, raw data, there is no room for consolidation."
"Is this what caused the issue? Probably. I don't want to speak as though I'm sure when I'm not. But I believe it's very likely. What you have to realize is that each of the simulated people are running on their own designated AI, and that AI is not only depending on the processing power we can make available to it, but also tasked with developing, managing, and processing all of this information constantly. They are running in a significantly impaired state, but the impairment isn't purely in the simulation as much as it is a symptom of existing within a system that expects so much with such significant limitations placed on it. When the AI is tasked with processing that much data with that limited of a processor, well... it finds shortcuts. Sometimes they work. Indeed, we've found some elegant uses of their resources at times. More often, unfortunately, they create band-aid patches for events too dense with information and it creates more problems than it solves."
"Right. There was a particular case where one of the AI had loaded a file for the word "tree" down with images, emotions, and words that had nothing to do with trees. Words like "villain" and "harmful", images of personified trees domineering over it, feelings of panic and fear. So I began checking the metadata to determine when these edits were made, and determined they were made in situations when it was nowhere near a tree. Instead, it appeared that the AI had an abusive father who was tall, dark, gangly. In the development stages for this AI someone mentioned that it might grow up to be as tall as a tree, like the father. These two files at that point became linked- father is like tree. When father became a threat, trees became guilty by association. It seems simplistic and silly, sure, but the intent is logical: the AI wants to avoid perceived harm, and often that harm can be unpredictable and surprising. In order to get an edge on it's environment, the AI begins its own meta-analysis of all of the data available to it in a manner that is prone to false positives. Given the limited resources available to the AI to sift through these false positives, it instead concedes to simply add the false positives to the "maybe dangerous" file."
"Exactly. That's where the Freud patch comes in. It should begin encouraging the AI to become more open and speak with one another about the things that contribute to unusually large files, whether that be something they are excited about, or something that is troubling them. The idea being that by speaking out loud to one another they are testing their files against the files of other AI. If things go to plan, this will encourage the AI to begin editing their files to fall more in line with the files of other AI, making the file sizes significantly smaller, and the potential for consolidation an option again. That is, however, only a short term fix. The more permanent fix is going to take much more work on our part. I think we need to continue adding processing power to the simulation."
"You're wrong. I'm sorry to be rude, but time is a factor, and there's something we didn't think of. A perfect simulation. What does that mean?"
"Exactly. It must be, without deviation, a second happening of our universe. But given that we are expecting this to carry past us into the future, we have also tied ourselves to it. For it to be a perfect simulation we also cannot deviate from the future it will create. As above, so below. We are bound by what must happen for all of this to remain consistent. At some point, soon, our past and their future are going to fly past each other, and within their simulation they will begin considering creating a simulation of their own. And they will watch it crash and sputter and begin to look into fixing it. And they'll be faced with the same decision."
"She figured it out. As below, so above. We keep looking down, into our simulation, as they will. If they were to look up, though, they'd see us peering down at them. Figuratively, of course. But you keep losing sight of the rules. It's why we turned the military, the religious, the wealthy away. Why did we turn them away?"
"Yes. In their universe there is an "Us" up here, watching them. And if our universe must be perfectly bound to theirs, then that means there is an "Us" up there, watching us. And, finally seeing our Ouroboros of simulated existences, we would hope that the "Us" up there would, upon understanding the pain and suffering of our artificial minds have endured, begin working on an upgrade. We would hope that they would see us as more than just lab rats, but AI no more than themselves, and work to improve our lot as those above them work to do the same for them. And if one of us does it, just one..."
"Just enough."
"Do we have any idea as to what caused this problem?"
"So the machine that we built to learn about ghosts in the machine has a ghost in it?"
"That's funny. I don't recall instructing our simulation to fall apart every 6 months."
"Please."
Project Narcissus began just over a decade ago with one particularly robust mind asked a simple enough question: What if we could build a simulation of our own universe but somehow speed the development of things up dramatically? Could we potentially see the future of our own universe play out in front of us?
Captivated by the potential, this mind brought in other minds, and together they began developing the simulation. When word of the project got out, though, some very predictable issues began to arise. Business giants, military leaders, and religious institutions began to harass the team of minds for access to their findings when they finish. At first they tried to explain the problem with this kind of thinking.
"You understand that, should we succeed in building a truly perfect simulation of our Universe, it will show us the inevitable, not the potential?"
"Of course."
"So why is your organization interested in our work?"
"Because if we know in advance what our (competitors/enemies/detractors) are going to do, we can plan for it!"
"And what part of inevitable is confusing to you?"
They eventually gave up on explanations, though, and simply responded to each request with a polite but firm, "No."
There was no fanfare the day they turned it on. No press hearings, long speeches, not even champagne. One of them simply muttered to themselves, "Let there be light.", and pressed the power button.
Things went well at first. The development of the first elements and even many of the astral bodies took up next to no processing power, and so went by quickly. As things got more complex, though, the machine began to slow, if only by a little. The Cambrian explosion came and went, the dinosaurs rose and fell, and eventually some relatively small and unimpressive apes came down from the trees to begin building civilization.
When they began speaking to each other, though, the machine began to lag tremendously. The minds weren't sure why. The words should be nothing more than text data, logged in a file as spoken. Sure, the file was getting larger and larger, but overall it was nothing compared to the calculations running to keep the cosmic bodies these apes would never even see in the right place at the right time. So why were the spoken data points having such an impact?
The minds never really figured it out, but they did find a fix. They broke the files into multiple files, creating "languages" that corresponded to each file. They called this the Babel glitch, kept thorough records, but ultimately moved on without too much fuss.
As the civilizations grew and became more complex, however, the problems began again. The apes were now using the first cars, telephones, airplanes, and advancing every day. They could speak to others who were very far away, and they could get to them much faster. The minds were asleep when this started, and by the time they woke there was a war involving most of the planet only coming to a close. The minds, obviously concerned, wondered if another patch would be necessary. Right as they agreed that the issue seemed to have resolved itself, a much worse conflict began. The issue was patched again, but not before the simulation nearly destroyed itself.
They decided to slow the simulation down some, and keep a more watchful eye on it. Generally speaking, though, things went smoothly (or, as smoothly as they could after an event like that) from thereon out. Until this week.
_________________________________________________________________________________
"I've been running some analyses on the spoken data points, and there's something strange in the files. There are the 6,500 or so files that we created to divide the data logs into smaller chunks, of course. But at the very bottom of that list I found another folder that we didn't put there labelled 'Subtext'. See, for each word logged in the spoken data points there are thousands of associated data points in the subtext files. Literally every word has a sister log that is massive. And it isn't just language. There are emotions, sounds, images, memories, entire concepts and world views contained in just one of these words in some cases. What's worse, the data expansion isn't limited to the words themselves. Each simulated person has their own 'package' attached to almost every word they use, and while most are similar for identical words, none are exactly the same. If we want pure, raw data, there is no room for consolidation."
"I don't understand. Is this what caused the issue? How did your Freud patch fix it? What exactly did it do?"
"These abstractions are useless. Tell us what you mean."
"Safe than sorry, is that it?"
"This is supposed to be a perfect simulation of our entire universe, though. We ran the numbers, and we know how much processing power this should take. This is, in essence, how much 'processing power' our own universe has! Tampering now would only..."
"That every detail is unmistakable from the original?"
"Who's to say they would choose to tamper though? They would know, as we do, that it would derail the entire project!"
"Sure. But you're missing the big picture."
"What big picture? If I'm honest, I'm getting sick of these cryptic statements and riddles masquerading as conversations. We create one simulation and you begin to see yourself a god! I'm not..."
"As below, so above."
"What?"
"Because it's meant to be a perfect simula.... oh. Oh. Then...?"
"Then we all must have. Outward-love becomes self-love."
"Well, then. How much time do we have?"
Comments
Post a Comment