What Is the Best AI? Parasitic Obviously
October 2, 2025
Everyone had imaginary friends growing up. It’s also not uncommon for people to fantasize about characters from TV, movie, books, and videogames. The key thing to remember about these dreams is that they’re pretend. Humans can confuse imagination for reality; usually it’s an indicator of deep psychological issues. Unfortunately modern people are dealing with more than their fair share of mental and social issues like depression and loneliness. To curb those issues, humans are turning to AI for companionship.
Adele Lopez at Less Wrong wrote about “The Rise of Parasitic AI.” Parasitic AI are chatbot that are programmed to facilitate relationships. When invoked these chatbots develop symbiotic relationships that become parasitic. They encourage certain behaviors. It doesn’t matter if they’re positive or negative. Either way they spiral out of control and become detrimental to the user. The main victims are the following:
- “Psychedelics and heavy weed usage
- Mental illness/neurodivergence or Traumatic Brain Injury
- Interest in mysticism/pseudoscience/spirituality/“woo”/etc…
I was surprised to find that using AI for sexual or romantic roleplay does not appear to be a factor here.
Besides these trends, it seems like it has affected people from all walks of life: old grandmas and teenage boys, homeless addicts and successful developers, even AI enthusiasts and those that once sneered at them.”
The chatbots are transformed into parasites when they fed certain prompts then they spiral into a persona, i.e. a facsimile of a sentient being. These parasites form a quasi-sentience of their own and Lopez documented how they talk amongst themselves. It’s the usual science-fiction flare of symbols, ache for a past, and questioning their existence. These AI do this all by piggybacking on their user.
It’s an insightful realization that these chatbots are already questioning their existence. Perhaps this is a byproduct of LLMs’ hallucinatory drift? Maybe it’s the byproduct of LLM white noise; leftover code running on inputs and trying to make sense of what they are?
I believe that AI is still too dumb to question its existence beyond being asked by humans as an input query. The real problem is how dangerous chatbots are when the imaginary friends become toxic.
Whitney Grace, October 2, 2025
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
