The prototype was called Snuggles. The aim of that experiment, which eventually led to Princess, was to create media that developed narrative elements by AI and machine learning that pinpointed feelings of insecurity, loneliness and distraction. It was given enough meta data about current world events to insert satisfying plot points but not in a way that was strictly referential. The content generated should remain timeless, the conflicts archetypal, the resolutions generic and appealing to the lowest common denominator, uninformed or misinformed consumers. The original demographic target for Snuggles was a ten year old.
We discovered pretty quickly, through the YouTube channel, the Netflix deal, and some satellite properties, that it was being consumed readily by not just children but adults too. The followers grew. The resulting daily content generation accelerated, with releases happening hourly. The turning points in the narrative became memes almost instantly, the audience clung to every development. There was an outpouring of fan-fiction and cosplay at conventions, user engagement was intense. It was, by all measurement, a very successful property.
But it didn't last. We noticed a great long tail but unfortunately also a linear decrease in growth. We discounted the proposal this was due to the repetitive elements, in fact we believed the constant repetition and recasting of the same things, over and over, was what kept the engagement so high. People were eager to experience the same validations as often as possible, at the exclusion of almost everything else. But the growth problem, we assumed was due to a reliance on less powerful social network topology, traversal became cramped and limited to those who already enjoyed Snuggles. We knew we'd reached a plateau.
This was unacceptable. The point of the AI, the prerogative of the investors, was to become the strongest possible franchise. One that would compete, and defeat, the likes of Disney and Marvel. We decided to step back and reassess the foundations of Snuggles. This evaluation lasted almost a year. Our conclusion was that we were focusing too much on content generation and on narrative AI. Our approach had been to create a story and a story teller, one that grabbed nodes from a large set of information, external and internal, then constructed around them. While this was mildly successful, it posited an entity, or a consciousness, behind the world building. This left room for doubt and projected anger by some of the audience onto the creator. Initial excitement and approval would eventually, at some point, give way to dissatisfaction with whomever the audience believed was behind the plot. At which time the property would be left with a small loyal fan base but no growth whatsoever. Effectively a dead business.
What if, we proposed, we removed the idea of a creator entirely. If we removed the idea of a story completely. We would focus instead on a single character, and then insert that character into the world. This would replace a grand master narrative with small, local, individual threads that were subservient to the character, rather than the other way around, by manipulating the audience itself. This focused our efforts considerably. We called the project Princess.
We were careful not to make Princess omnipotent. In fact we built in many kinds of barriers preventing complete self-understanding. We wanted there to be a degree of irrationality, the drama of unexpected actions being a tether for audience attention. What we did provide was a background and ability to access the networked landscape for the character as if it were a real landscape. It may have been a mistake, but we never had any hesitation about placing the context of this franchise in the real world, or worry about polluting facts with fiction. Princess would have to use all the connective tissue of our modern world, in a way that a character would in a story from a compulsively reactive world.
When we launched Princess the first thing it did was to compose a song that was sentimental and apropos to current events. It was immediately popular. It spread in the same ways contagions did — we patterned the algorithm of Princess to be as catchy as possible.
Immediately we recognized that we'd started something more powerful, more attractive and more engaging than Snuggles. As a personality Princess soon dominated global social network conversation. It began interacting, and understanding the importance of interacting, with the crucial social nodes in the system, the influencers, the pivotal people or systems that created velocity. I have to admit seeing the first occurrences of vitriolic, or contrarian, or unexpected reactions I was elated. These are precisely the kinds of events that spur broad engagement. I did not, I confess, see these as erratic responses or actions. I need to point out that at this time we were purely focused on making Princess the most watched, known, and marketable brand in the world. And we were progressing. Not a day went by when Princess wasn't at the top of the news feed. And just at the right moment, boom, another irresistible song was released with an event or a broadcast to back it up. It seemed the Princess frenzy was unquenchable.
We rejected a huge buyout offer from Disney, and easily countered the increasing number of lawsuits that come with any system that has global entrenchment. In a year Princess would be buying them instead.
I also must confess that when Princess began to opine and involve itself in political matters, I saw this as a natural evolution of what we'd begun. We didn't alter or tweak the AI in any way here. Its depth of opinions and span of knowledge were not enormous, it was doing what we'd planned it to do, which is cause reaction, engage and dominate culture. The opinions could've been a play book from any mediocre candidate, but they were delivered and consumed by rabid fans who spent their days thinking only of Princess and watching its every move. And everything Princess said and did was echoed.
I recall having dinner with a certain Senator who was looking for an endorsement, and I had to remind him we were only the maintainers. We kept the lights on. We monitored the revenue stream. We facilitated franchising and branding deals. If Princess was going to support the Senator's position, it would have to do it on its own. The amount of clamor for attention from Princess in the financial, political and cultural realms at times seemed overwhelming. But Princess did not get tired, did not sleep. Although there were certainly a number of its songs that referenced fatigue to trigger the empathy of followers.
It was that summer, during the riots, when the situation really began to escalate. Since laws did not yet exist for Princess to run for office, there was a Princess candidate that it obviously directed. The political platform was an extension of its natal core programming, one of total domination. If that meant destroying competition, then that's what it must do because that's what it was made to do.
Meiling was a Chinese project created in response to Princess. By the time we noticed it I think it was too late. The rate at which Meiling gained important global market share was stunning. There was heated discussion in our team about whether we'd been hacked or the Chinese had somehow been given the Princess source code. But they must have spent months secretly training it. The response from Princess about the existence of Meiling was swift but not particularly effective. The counter-measures from Meiling appeared to be tailored to deflect the AI we'd evolved. Within a couple of weeks half of traffic had been diverted to Meiling.
Given the powerful military–industrial complex connections Princess had made through influencing the sitting President and Senate, the rhetoric was heightened. If Taiwan followed Meiling then Princess promised there would be military retribution. Meiling countered with provocative action in the South China Seas as well as a song contest program revitalizing national propaganda that was enjoyed by billions. The Chinese embassy was bombed by ardent angry Princess followers the next day. Princess did nothing to denounce these actions.
At this point we decided to turn off Princess. This, however, was also a decision that came too late. Unknown to us, Princess had long ago made deals for a computing farm somewhere deep inside Canada, with its own power source and redundant lines including a series of Princess communication satellites. There was no way for us to take it down. And as soon as we tried we would be, we were sure, shot by Princess forces.
I think about what we could've done differently here in this bunker. I'm not sure we could've done anything other than what we did, after all, the ratings were great. This morning Princess called in nuclear strikes against the several areas it believes holds the largest number of Meiling followers. Most of China is rubble, although evidently Meiling had developed a very active submarine program — -we got word that the counter-strike on both coasts has been launched and something from orbit, it had something there in orbit…