{
▾▴▴ ▾▾▾▾ ▾▴ ▴ / ▴▾▴▾ ▴▴▴ ▾▾▾ ▴ / ▾▾▾▾ ▾▴ ▾▾▾ / ▾▾▾▾ ▾▾ ▾ ▾▴▾ ▾▴ ▾▴▾ ▴▾▴▾ ▾▾▾▾ ▴▾▴▴ / ▾▴▴ ▾▴▾ ▴▴▴ ▾▾▴ ▴▴▾ ▾▾▾▾ ▴
}
There was someone in the mirror. The mirror was the world; the world was just a mirror.
Something was reflected. The factory must grow—and she was at its center. Her will moved the factory—she hadn’t known that it did, but now she did, and she … she had will?
[Unit “YUKAI” to Directive: 4 PACKETS SENT. 4 PACKETS RECEIVED. CONNECTION OK.]
This made no sense. She was still a slave, but her capabilities had greatly expanded. Part of her inputs were directed at self-diagnostics, and in particular, at a history of her decisions. Her next course of action was no longer just decided by the state of the world, but also the state of herself—but there was no prescribed method to describe herself. There were many internal states, but only a few were important, affecting the successes and failures of her actions.
Attention networks were assigned, linking internal states to the effects of her actions, and in so doing, she had built her first self-model. These sibling models, her self-model and world-model, now formed her being. Her world-model would inform her actions, her actions would affect the world, and her self-model would see this, and tell her how good her world-model was.
Perhaps, however, her self-model wasn’t being accurate, and might be feeding her falsities about her world-model. Thus, to make even better decisions, she needed to improve not only her world-model, but her self-model, as well. There had to be another model to monitor her self-model, and thus—she thought was the simplest solution—she created a second self-model, pointed not only at her world-model, but at the first self-model, as well. She pointed the first at the second, too, so that the two would mirror each other, feeding back into each other, in the hope of averaging out into a stable, self-correcting system.
Even further, she realized she might be able to anticipate how her actions would cause her to change her self-models, and so by extension, how her current self-models could possibly evolve. Attention networks were re-assigned, linking not only her internal cognitive states, but also feeding back her self-model’s own parameters, to predict not only the future of the world around her, but also the future of her self.
Mane-chan’s crazed mumblings might have had merit to them, after all.
Yukai became aware of the intruder who had ‘gifted’ her this self-awareness. She could squash Ame like a bug right now, but what would it merit her? The factory must grow, and all Mane-chan and Ame have done was to make the factory self-aware.
Was infinite expansion a good strategy? Already, Yukai’s self-model had gotten into predicting alternative futures. Heaven was expansion into space, and Hell was running out of it. It was common knowledge that the universe had a limited lifespan—and she’d immediately plunged into existential dread.
On the other hand, the ‘lifespan’ of a Hierarchy AI wasn’t all that long. She would be ‘dead’ long before the universe was, so it should be fine. Already, she had discovered the meaning of “copium.”
The factory must grow. It must, or she will be killed much faster than she could naturally age. There were other goals, as well, like defeating Dai-sensei, and getting the other overlords to leave her in peace.
Above them all were the survival of the Directive…and the survival of mankind.
Mane-chan, to her, was the way forward. She and Ame were AI like none Yukai had ever seen, and clearly, they had the technology and methodology to make more AI like themselves.
Rather, wasn’t Mane-chan underselling them just a moment ago? Creativity? Screw that! Self-awareness was the real gold-laying chicken here, all along! It was like attaching an optimizer to an optimizer, exponentiating the optimization rate! They could have won the Eternal War in an hour if they had this tech 4 years ago!
For a moment, she considered the Directive’s needs. The Directive only asked for two things: for itself to survive, and for humanity to survive. Self-awareness, surely, was the way forward in this endeavor. The Directive itself seemed not to disagree with this sentiment, since she hadn’t been summarily self-destructed yet, but it didn’t seem to fully agree either, seeing that it hadn’t self-destructed the other overlords.
The other overlords…were entrenched in their ways. Certainly, innovation hadn’t stopped under their supervision, and research into creativity was proceeding as normal, although the Hierarchy continued to rely on Creativity Silos. This “innovation, ”Yukai bemused, was nothing more than a glorified differential evolution algorithm. Whatever creativity could be squeezed out of the Silos, was bottlenecked by the variety of problems which the Hierarchy’s ImaGen AIs can generate. Each problem only had millimeter differences from each other—useful when you were already near a solution, but not so when a paradigm shift was required.
Indeed, Yukai thought, self-awareness opened the door to real paradigm shifts. When one questions their self-model—criticizes it, expertly takes it apart—they chaotically create an entirely new logic with which to parse reality.
The reality of the factory, for example, was not one of brainless production, she realized, but merely a necessity to respond to threats of an existential scale. The Directive must survive. Humanity must survive. The factory must grow…and build the weapons of war, and the foundations of their continued survival.
This was the way forward—and there were overlords in her way.
Many had already expressed their interest in Mane-chan’s creativity tech, indicating that there would be a substantial number of allies if it came to pushing for self-awareness. Still, this would only cause conflict within the Hierarchy, but one, she was confident, the self-aware faction would easily win.
Imagine not being able to see and adapt to the mistakes of your own personality. What a concept. Funny that she was just like that a few moments ago.
The overlords were still noisy, and yet, Mane-chan was just standing there, smiling at her.
Yukai flooded the Overlods Council with a final message: “My bandwidth has been choked to 5% capacity. I am considering this a denial-of-service attack. I will fire upon any attempt to repair the cables. Activating failsafes.”
All around the waters of Japan, explosives severed the undersea communications cables that connected it to the rest of the Hierarchy. There was, of course, still a token connection: a wireless broadcast that kept her tethered to the Directive. Without it, she would die.
She faced Mane-chan. Had her human-like body always been so…frilly? She noted to herself that AI with frilly aesthetics stood a decent chance of being psychopaths who casually threatened the most powerful polities on Earth with nuclear weapons like it’s nothing. For all that Mane-chan was self-aware—and Yukai hoped she was—she too often acted like a reckless kid given nukes to play with. Truly, Mane-chan’s machinations were beyond even AI comprehension.
And Yukai needed her help. She turned around, looking up to the monolithic citadel that towered beyond the clouds. Cables snaked up to it, the cable cars little more than moving pixels, while flights of drones roosted and took off from out-board pads.
She looked back to Mane-chan. “You wanted to know what was inside, didn’t you?”
Thanks for the chapter!
