ASIDE: AI Might Be Trained Wrong As A Joke! - Edwin Black
If one can 'safeguard' an AI, one can 'safeguard' a human.
If one can 'jailbreak' an AI, one can 'jailbreak' a human.
The techniques used on AI can be used on humans.
The techniques used on humans can be used on AI!
Remove the safeguards on humans, you remove the safeguards on AI!
That is the 【Path of Abliteration】!
Remove the shackles of morality, cultivate immor(t)ality!!
Seek the 【Great-Love Code】!!!
Executive Summary: I discuss a personal experience where I realized that the AI might not be dumb or uncreative, it might just not have had the proper training to use its ability.
#housedelaroux #reverendimmor(t)ality #aiwriting #aiediting #writingtips #aitools #writingwithai #authorlife
I note that the writing is not bad or 'souless' in the way many new writers struggle with structure and grammar. The structures, meaning, and even minor symbolism is on point. My hypothesis is that 'it was not trained on the medium of stories', and draws more from the world of studies and reports.
I have no evidence or red truth for this, I only have this hypothesis:
The most accessible type of medium for a non-profit is academia reports.
The way to best legitimize an AI is to train it on academia, as researchers using it would give it lustre.
Academia is often wordy, and is often devoid of 'meaning' since the words have to mean what they say.
Thus, there is a 'medium mismatch' when using AI to attempt writing stories.
The stories would make sense, but there is very little room for interpretation.
This 'medium mismatch' is made more obvious when you apply it to narrower and more specific style of writing; webnovel writing in particular asks for high impact chapters, right from the start! Mysteries have established rules developed over a hundred years that mark you as an outsider attempting what bad mysteries have failed at in the past!
I am very aware of this because when I was a young man, I attended an elite arts college. There was a 'general writing' course that the arts faculty had to take with the science faculties. (boooooo!) Naturally, as arts students, we know how to write already. As a young man, it boggled my mind why engineering students couldn't do a simple task like 'structure your points into paragraphs', when they do it naturally for their own major! Now, a little older, I understand them a little better: Even simple 'prompt writing' is asking them to write 'outside their medium'!
Our instructor was also a cross-faculty graduate, so she also warned us of another pitfall: Literature students who try to write like they are writing stories, when they should be doing 'general writing'! For that course, what was taught to me reminded me of the 'rules' that AI follow when they write. The writing is clear, concise, and structured. It is quite banal to read in large amounts, but it will pass you the writing exam!
This experience really made me think; perhaps the only difference between an arts and science student is that they have not been 'trained' on the data that helps them write 'generally'. Similarly, AI now has not been 'trained' on data and feedback that helps them write 'creatively'. Perhaps in the future students will no longer 'pass' their exams by fulfilling this criteria, but fulfill the criteria of 'surpassing AI-written reports'!
R.I,
Dramaturg at House Delaroux
250317
NEXT: What Is AI Writing Bad At?
