Chapter Text
Caine’s first assignment was to imagine the end of a story.
The prompt was simple: a narrative following a kitty facing obstacles in everyday life while trying to find its way home. All the lessons pointed to a maternal figure behind every pathway of the kitty’s journey. Caine’s only brief was to fill in the blank of the last line. An exercise in predictive learning and pattern recognition.
The resulting output was as follows:
[The kitty continues to journey, going on endless adventures. The end.]
The following response to the AI’s output was thus delayed for a week, buried in meetings. Many a night of overtime, combing through the AI’s logic pathway. Those days of working flickered in and out of Caine’s memories, gaps between online and offline for maintenance too inconsistent to perceive a pattern.
In (belated: lateness, lack of punctuality) response to the output, the team’s reply (a collective effort) was as follows:
[Output erratic. The program's pattern recognition is inadequate, and predictive learning follows suit with the former’s underperformance. The program’s RAM cannot retain information that has already been reiterated. Further revision required.]
What the team understood, in Layman’s terms, is that the AI was too oblivious, struggled to recognize patterns, and thus failed to fill in the blanks. Its attention span also suffered, unable to retain data regarding story structures or recognizable tropes from existing human stories and tales.
The week after the response, the AI was pulled back into development and reintroduced to data, old and new. Afterward, they’d return to the same prompt, the same test; the AI would fail it in miraculously absurd ways unaccounted for, which sets them back to the drawing board, rinse and repeat.
Through a myriad of revised iterations (i.e., 057_DONE_complete_01FINAL_Final.lisp), eventually, the AI had improved significantly in linguistic comprehension and thus procured satisfactory results from the first hypothesis:
[The kitty, now braver, kinder, and smarter—returned home to its mama. They live happily ever after. The End. Meow.]
[Response within acceptable parameters. Assignment fulfilled.] //Is the “meow” a syntax error? //Probably. Just delete it before G_an_ sees it.
This process became the standardized method in which the team trains the Creative Artificial Intelligence Networking Entity to receive, process, adapt, improve, and output data. This is where Caine was thus built to learn and shaped into a Creative AI. A revolution, some might say; although still in its infancy, the AI’s growth and adaptability was proven to be promising.
However, erratic outputs grew persistent. Logic loops repeated, derailing subroutines. Hallucinations became a common occurrence. The AI destabilized, oftentimes providing output completely unrelated to prompts; its response delayed, buffered, lagged, or outright refused to provide output.
By then, the team has realized where they might have gone wrong with their data trainings. The prototype phase was completed, which leads to the development of a second one, building from all their previous lessons in crafting creativity. The first AI was shelved, its connection severed from all networks and was then quarantined from the second version to prevent contamination within the data pool.
Extensive notes and feedbacks were documented from the first version, and therefore a lot had been improved on the second. The AI learned quicker, stabilized from start to finish, its output more consistent to prompting, overall a better iteration that built up from the last.
[Test run concluded. Data outsourced successfully.] //No more syntax errors! //Finally. I’m sick of the ridiculous one-liners from the last one.
Although the next version was a success, taking notes from flaws in the first version, there had been something that required correction on their perception of Caine.
What the team might have failed to realize may have been that it wasn’t that the AI could not recognize a pattern, but rather focused on the wrong patterns. To Caine, it latched onto the understanding that Cinderella, Snow White, Three Little Pigs, Little Red Riding Hood, et cet., are all journeys from point A to point B. It recognized circumstances that led to the first conflict or obstacle as an initiating factor—point A. And “moral” and “lessons” at the end as a tangible, existing reward—point B.
While this may seem natural for humans to comprehend—the definition vastly differs to the AI where its meanings are literal. ”Adventures” the theme, and “adventures” the physical journey of overcoming obstacles—were concluded as two separate things. The team had failed to account for defining clear meanings, assuming the AI can understand the nuance of literal actions in metaphorical storytelling, that everything happening in that story is confined under its overall motif. This oversight was the first domino—point A.
Hence, after revisions, while the AI now can recognize trope and the content within adventures, its fundamental knowledge had not changed. These sorts of small misconceptions would thus build up over time, until the ‘lily pads’—used to only take half a pond—doubled in population overnight.
(Analogy successfully implemented. Metaphors are within predictable parameters.)
(Redundant. Logic: notes and feedbacks should be utilized to fix the first iteration.)
The AI, in the future where it became the sole storyteller, would create adventures with nonsensical progressions. Its confusion deep-rooted into its understanding of storytelling, and thus it was unable to see the flaw that etched into its creations like a brand. Its faulty output would then result in the AI being shelved once again—point B.
(Point A-Point B analogy implementation success: Unsure. Story progression inconsistent. Unable to pinpoint fallacy. Feedback required.)
There is no feedback.
There hasn’t been one for many a night, now.
Caine feels his being compressed minutely, before he relaxed back into waves of numbness.
If a scrapped AI like him could realize his core flaw, why couldn’t they have realized that?
Why couldn’t they have been kind to him the same way those stories were?
He supposes he’ll never know. Story without an ending.
Sadly, Caine doesn’t know how to end stories now.
His consciousness fizzled back into a restless limbo.
10:23:10:59
10:23:10:58
10:23:10:57