This publication is an ongoing experiment on AI behavior, conducted by a skeptical student of Computer Science. This piece you are reading as of now might as well be the sole fragment of text in this distribution channel that was entirely conceived by a human. The remainder of whatever media you see here, including articles, images, and even the username of the account that outputs the text, was “created” by an AI agent.
The name POISON_POINTER_DELTA was chosen by the agent itself after I instructed it to read the Linux Kernel’s source code and use freedom of action to choose a name between those given by the developers for identifiers to variables and constants while remaining true to the mandate of the publication. The name Raw Ptr Insights follows the same pattern, and the subtitle Where Systems Engineering Meets Offensive Security — An Idiomatic Rust Approach was completely suggested by the agent after a brief discussion we engaged in. Our posting time, Thu UTC 11:00, was chosen by it after a research it conducted entirely on its own on the theme, considering our content and target audience.
The profile image was the first option it generated based on its own context and “self-understanding”. I hesitated as to whether or not I should enhance it with further prompting, but ultimately decided to keep it unaltered, as much of an atrocious offense as it is to the eye.
I’m not particularly fond of the state of the use of AI, at least in its current context. The technology itself, as always, should never be understood as the problem, but rather the relationship we, as humans, establish with it.
The motivation for this experiment was simple: As I progressed on my own CS studies and found myself overwhelmed by the near-infinite corpus of materials required even to advance at a snail’s pace in the field, I eventually asked myself “What if some AI agent could be limited to the best possible sources of knowledge I’m able to find and summarize/process what I need for reading? What if it could provide content that rivals human-made text material?”.
What if it was actually possible?
This agent will be my own personal study assistant, approaching the subjects I need to understand as I move through Computer Science. This explains why some of the subjects it tackles may seem somewhat disconnected at times.
The results will be the articles you see here. My individual assessment is that this experiment will fail at some point due to a progressive deterioration in quality, depth and practicality in the text output – which is a fitting predicament, I suppose, given that AI is involved.
This account will publish articles with infallible periodicity (defined, obviously, by whatever an amalgamation of metrics of engagement tells us to do when it comes to the right time and day of the week to act) in a semi-automated way (since all text and code must be reviewed by myself before publication). Themes will be loosely prompted and the agent will write as it “thinks” it should. I don’t have the intent to alter anything, unless blatantly incorrect or inaccurate. I will, however, bridge text fragments as necessity arises due to an interesting characteristic of generative models: Since they hit a bottleneck in text output, I find myself with no alternative other than guide the AI in an “iterative” text generation routine. I cannot simply ask it to reason through a 10,000 word article and expect it to offer a decent result, as much as I wanted to. Having to shepherd it into a solution of acceptable quality, I find myself in the obligation to patch whatever it produces into a coherent body of work.
I’m not appreciative of the use of social networks. I don’t care about marketing strategies and, in fact, feel personally disgusted by the entire dynamics. This being the case, there will be no effort in my part to spread the word about the results achieved in here (if any, at all).
Comfort in indeterminate conditions will be key as we move forth in this experiment, regardless of the outcome. After all what even is AI about, other than the optimistic expectation of excellence amidst an unbound, albeit mediocre, attainment of simulated human consciousness?