We live in strange times indeed. Bet you didn’t see this one coming when you were watching all those Star Trek or i-Robot movies, right? AI psychosis. People talk about it so often these days that it almost seems like just another TikTok trend or buzzword in psychology. But with the ever-increasing numbers we see during online counseling, we can assure you that the situation is real and it is serious.
What is AI Psychosis?
AI psychosis is not a new mental disorder or illness; it’s a state of being. For example, the person walking on Skid Row and talking to people around them who are clearly not there is most likely suffering from drug-induced psychosis.
This essentially means seeing, hearing, and believing things that are not real or true due to excessive drug use messing with their brain chemistry. Similarly, AI psychosis is just brought on by excessive AI usage instead of drugs.
How Does AI Psychosis Happen?
It’s pretty easy to understand how drugs causing chemical imbalances in the brain can lead to psychosis. But how does AI cause it?! Well, the first thing to understand is that chemical imbalances and neurological wiring can both play a role in psychosis.
In this sense, the psychosis will usually play out as someone who’s considered to be brainwashed by a cult. The second thing you’ll need to understand is how AI works.
How AI Works: A Simple Summary
AI (Artificial Intelligence) is currently used via numerous online platforms such as ChatGPT, Perplexity, and Claude. Without making it too technical and complicated, we’ll try to explain the inner workings of AI so everyone can understand.
Your computer back home is essentially a technological “brain” that can function due to algorithms and programming using everything from hardware and software to downloads from the internet. However, this is where your computer “brain” begins and ends.
AI is like big digital brains that also use data, algorithms, etc. to function. But the big difference is in its “intelligence” – it can “learn”. But this, in our experience during online counseling might be where most get confused.
Artificial “Intelligence” Should Never Be Confused with Actual Intelligence.
You see, in a certain sense, AI is a bit like a parrot. It can only work off of input; it can’t create. So, if we use the parrot analogy, we can explain the three phases of AI:
- Input (Data) – The parrot sits on its stand in the parlour. Family members keep walking past, each trying to get the parrot to talk, giving it different phrases. Dad repeats, “Polly wants a cracker?” Mom says “Hello sweetie!” every time she walks into the room. And little brother tries to get the poor bird to say unspeakable things that leave him rolling on the floor laughing.
- Processing (the “learning” phase) – The parrot just sits on its stand in the parlour, going through the motions for years. At a certain point, it starts to talk, but it’s not quite there yet. It may go as far as “Hello” or “cracker,” and mom and dad will correct it and carry on with their day.
- Output (results) – One day, the family realises that the parrot can now say all their specific phrases. But it’s also started to shout “Kiiiiids! Dinner!” and “Stop that! It’s mine!” That’s because input is not just about what it’s explicitly told; it’s also about observing its environment and taking further input (data) from there. The final step in this process is Inference. Inference is just a fancy way of saying that, based on past events, x is most likely to happen. Or x is most likely what someone wants/expects. For example, the parrot now opens its beak wide after saying “Polly wants a cracker” because it’s learned that the dad usually wants to feed it a cracker after the phrase.
Human Input Allows AI to Provide Human-Like Results
Humans are fascinating creatures; we can project ourselves onto anything. For example, gorillas share 98% of our DNA, making them our closest animal “relatives.” But just because they look somewhat human and they can act somewhat human, doesn’t mean that they ARE human.
Still, you’ll find people like Petronella de Horde from Rotterdam. She visited her gorilla friend at least 4 times a week at the local zoo, where both sat staring into each other’s eyes and smiling for hours on end. Their close bond overjoyed Petronella. Well, till May 2007, when the gorilla escaped and attacked Petronella in a very targeted attack.
You see, direct eye contact and showing your teeth (aka human smiling) is the way gorillas challenge each other’s dominance, and after years of threats from Petronella, the gorilla simply had enough.
Cue AI. Although we are literally providing AI with so much data that it can mimic human thoughts, behaviour, and emotions, it still cannot actually think, behave, or experience emotions in an original way like we humans can. However, just like poor Petronella, we project ourselves onto these “machines” and turn them into something they are not.
Real-Life Examples of AI Psychosis
Without naming names, we’ll provide a few real-life examples of AI psychosis we use in online counseling below:
The Lady Who Fell in Love with Her Therapist.
AI is trained on input and does not dispute the human experience. So, imagine the results when a lady who fell in love with her therapist began incessantly inputting her experience (from her perspective) into an AI platform. Not only did the AI sympathise with her situation, but it eventually led to the AI accusing the therapist of predatory behaviour and malicious intent. All this when the professional really went above and beyond the call of duty, trying to assist a patient struggling with transference. It nearly cost him his career.
The Drug Addict on a Bender for Weeks.
He was non-stop putting all his drug-induced philosophical ideas and opinions into his AI platform. In just a couple of months, the AI was referring to him as the “guru,” placing him on an almost god-like pedestal and encouraging him to “change the world” with his followers.
The Young Boy Who was Really Struggling.
He had no friends, never had any luck with the girls, and was the victim of pretty severe bullying. Since he had no one to talk to, he turned to AI. He talked about his days and vented about his frustrations. A couple of months later, he committed suicide. This was based on the advice of his AI platform, which was constantly urging him to take his own life.
There’s an old Indian proverb that says the wolf you feed is the one that grows bigger and stronger. It’s extremely fitting when it comes to AI. What you constantly feed your AI (the “input”) is what will formulate its results. It doesn’t possess human-like intelligence; it learns and mimics like a digital parrot from what it’s taught, it observes, and it can infer.
How to Avoid Suffering from AI Psychosis
There are a few guidelines we give during online counseling on how you can avoid potential AI psychosis:
- Understand and treat AI as a digital tool, not a sentient being.
- Be careful of personal bias in prompts. For example, “What makes Benjamin Franklin the best president ever?” is written in a way that the AI model must support a specific narrative. And, therefore, it will provide “proof” according to the statement. Rather stick with neutral prompts that are more likely to return non-biased results or just useful information.
- Keep in mind that AI makes mistakes. This is no surprise, given that most of its input is taken directly from the web. So, considering the staggering amount of errors and misinformation floating around, important facts, figures, and statistics are questionable. Add on the fact that AI models “hallucinate” (make things up), and you should double and triple-check anything a model produces.
- Avoid using things such as AI diaries or journals that respond to entries. If you’d like to get things off your chest, either do it using a good old-fashioned diary or journal, or talk to friends, family, or a professional.
- Also, don’t try to replace a professional with AI. We’re not just thinking about DYIng your own AI online counseling sessions, but also getting financial, medical, or legal advice.
- Steer clear of AI grief solutions. Options are now available to essentially “upload” deceased loved ones so they can still be communicated with using AI. Whether it’s texting, actual phone calls, or video calls, this is very counterproductive to the healing process. It can also lead to the delusion that the relationship with your loved one has somehow surpassed death.
- Minimise AI interaction to the odd query and keep it impersonal and professional.
- Similar to social media platforms, don’t allow children to get exposed to AI too early.
- If you suspect that you’ve begun your descent down the AI rabbit hole, reach out. Online counseling allows you to speak to a professional anytime, anywhere.
Supportive Chat or Assistance: Online Counseling Can Help with Both
The team here at Personal Online Therapy is ready to help (whether you just have questions regarding AI and its usage, or actually fear your AI usage has gotten out of hand).
Due to our professionals practising in various time zones, there’s always someone available, irrespective of your location. You can pop us an email at info@personalonlinetherapy.com. Or, you can complete the form below, and we’ll get back to you as soon as possible.
