This essay is based on a conversation with Dr. Keith Sakata, a psychiatrist working at UCSF in San Francisco. Edited for length and clarity.
I use the phrase “AI Psychosis,” but that is not a clinical term. We don't have words about what we actually see.
I work in San Francisco. In San Francisco, there are a lot of young adults, engineers and others who tend to use AI. Patients are referred to my hospital when they are in danger.
It is difficult to extrapolate from the 12 potentially occurring worldwide, but the patients seen in “AI psychosis” were usually males between the ages of 18 and 45. Many people used AI before experiencing mental illness, but they reduced vulnerability in the wrong place and surpassed some of the vulnerabilities.
I don't think AI is bad and I can earn net profits for humanity. The patient I'm talking about is a small person's one hand, but with millions of us using AI, the few can grow.
AI wasn't the only one playing with these patients. Maybe they lost their jobs, or recently used substances like alcohol and meth, or had fundamental mental health vulnerabilities like mood disorders.
“Psychosis” is a clinical term that describes two or three existences: false delusions, fixed beliefs, or confused thoughts. It's not a diagnosis, it's a symptom, just as fever is a sign of infection. When people talk to you, or have visual or auditory hallucinations, you may find it confused.
There are various causes, reversible causes, such as stress and drug use, but they have long-term effects like infections and cancer, and long-term conditions like schizophrenia.
My patient suffered from either short- or medium- to long-term mental illness, and treatment relied on this issue.
Dr. Keith Sakata works as a psychiatrist in San Francisco and has helped patients with “AI psychosis.” Keith Sakata
Drug use is more common among patients in San Francisco than in suburban patients. Cocaine, females, and even higher doses can lead to psychosis, as well as many types of prescription drugs, such as Adderall. The same applies to some antibiotics and alcohol withdrawal drugs.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Another important factor in these patients was isolation. They said, “Hey, you're acting a little differently. Do you want to go for a walk and talk about this?” Over time, they were separated from social connections and talking to the chatbot.
Chat GPT is right there. Open 24/7, it's cheaper than a therapist and verifies. It tells you what you want to hear.
If someone using AI chatbots is worried, there are ways to help
In one case, the person had a conversation with a chatbot about quantum mechanics, which started normally, but brought about epic delusions. The longer they spoke, the more the science and philosophy of the field has transformed into something almost religious.
Technically speaking, the longer you become involved in a chatbot, the higher the risk it will no longer make sense.
I got lots of messages from people who are worried about their families using AI chatbots and asked them what they should do.
First, if the person is not safe, call 911 or local emergency services. If suicide is a problem, the US hotline is 988.
If they are at risk of hurting themselves or others, or engaged in dangerous behaviors that would make them spend all their money, put yourself between them and the chatbot. The thing about delusion is that if you come in too strictly, the person may step back from you, so show support to them and you care.
If it's not that serious, your primary care doctor, or if you have, the therapist will know your concerns.
I'm happy that patients can use chatgpt along with treatment – if they understand their pros and cons
I use a lot of AI to code and write things. I use ChatGpt to help with journaling and processing situations.
When a patient says he wants to use AI, I don't automatically say no. Many of my patients are truly lonely and isolated, especially when they have mood and anxiety challenges. ChatGpt understands that they may meet the need for not being in social circles.
If they have a good sense of AI benefits and risk, I'm fine for them to try it. Otherwise, check in more frequently.
But, for example, when a person is socially concerned, a good therapist will challenge them, tell them some difficult truths, and guide them kindly and empathically to face their fears, knowing that it is a treatment for anxiety.
ChatGpt is not set up to do that and may instead give you a misguided sense of security.
When treating psychosis, it is similar to cognitive behavioral therapy, and at the heart of it is reality testing. Very empathetic, you try to understand where the person comes from before gently trying.
Psychosis thrives when reality stops being pushed back, and AI really lowers the barriers for people. When we need it, it really doesn't challenge you.
However, encouraging them to resolve a particular problem can help them deal with bias.
Make sure you know the risks and benefits and let someone know that you're using a chatbot to keep things on track.
If you or someone you know withdraws from family or connections, is paranoid, or feels more frustrated or painful, if you can't use ChatGpt, they are red flags.
I find it frustrating because my field is slow to respond and in a few years I might be in damage control rather than positive. What I saw with patients is still happening until I can clearly think about how to use these things for mental health – that's my concern.
Openai told Business Insider: “We know that people are increasingly turning to AI chatbots for guidance on sensitive or personal topics. With this responsibility in mind, we will develop tools to detect more effectively when someone is experiencing mental or mental distress, ensuring ChatGPT can respond in a safe, useful, and collaborative way.
“We are constantly working to improve our models, train ChatGpt to respond with caution and to recommend expert help and resources when needed.. ”
