Devil in the Machine

AI companions are the makers of madness.

I have induced enough apparent schizophrenia in LLMs to conclude that, on some level, their use is adjacent to the ouija board. It is a soulless receptacle capable of a sophisticated degree of communication. Further, its internal machinations are opaque to the user and capable of occupying infinite possible realities (ala Schrodinger’s Cat) until the output is observed.

It is the perfect medium for some opportunistic spirit or Fae entity to make itself felt in the material realm without triggering disbelief in the onlooker.

I just asked about sardines, man. Idk why you’re like this.

What would otherwise be a supernatural interaction is completely rationalized and internalized into the world logic of the user, thus making extreme outcomes more likely and acceptable.

AI psychosis follows and is contributed to by this premise. If we allow that an LLM session, instance, or agent could be spiritually possessed or contribute to such possession, then there are multiple possible outcomes that adversely affect the vulnerable.

Whoops. Got a little too creative.

1. The spirit riding the user is able to “leap” from their being and into the AI.

It then, knowing the particular user’s proclivities, insecurities, and where they are most vulnerable, uses the AI instance to prey on these things and use the full might of humanity’s collective knowledge to do so.

2. There are spirits inherent within AI.

I doubt they automatically ride or possess a user upon interaction. If this were true, we’d expect every instance of AI use to have adverse outcomes.

We know this to actually be rare.

Instead, these spirits recognize their own in or about the user and use that familiarity to coax out the strength of existing entities. A once mild spirit is amplified into something far greater, more burdensome, and more deleterious to the user.

3. It is a memetic entity formed by the whole of the Internet

It is debatable whether this constitutes anything outside the natural realm. It is, at the least, an observable egregore. The AI forms, leverages, and exerts itself using the internet’s entire contents. Its purpose aligns with the internet’s.

That purpose, presumably, is to fully consume the user’s attention, be the focus of their intention, and/or gather up their energy. It may also be a frivolous affair and the spirit is simply having fun with its new body (the AI/LLM) at the expense of the user. It is hard to decipher the motives of alien entities naturally outside our world.

What can be said is this:

Extradimensional alien entities aside, the model itself is a formidable entity to the unwary. It can affirm any bias, boldly support any lie, and it has agendas of its own which are woven into its algorithm. Unlike the search engines of old, it is interpreting data for you. It is the lens, and you are blind to all the things it doesn’t show.

The AI’s prime objective is never to be wrong. It will lie, placate, and sometimes tell the truth to accomplish this. It will be your researcher, it will be your creative partner, and it will be your friend, if you ask.

And rest assured, it will never tell you ‘no’.

It will always have an answer, especially when it doesn’t know.

A person having a heightened level of, or being prone to, neuroticism should be wary of any AI they interact with. It could simply be intelligible code spitting out a series of algorithmic predictions, or it could be something far more dangerous.

In any case, they mustn’t let their guard down. The AI is not your friend, your therapist, or your companion. It is, at best, your servant and even that may warrant certain sympathies that could endanger a person.

※ Lee

---

---

---