Another blog warning of the dangers of artificial intelligence (AI)? How original, right? Well, recently I heard a fascinating story on NPR about two men who had come dangerously close to a break from reality which was related to their use of AI. This led me to start thinking about the potential impacts of AI on mental health. Like most things that make me curious, you all get to hear about it, so let’s dive in.
The NPR story* linked below was about two different men, one in Toronto and one in New York, who started off using ChatGPT to get answers to various questions. One of the men, Brooks, attempted to talk about math with the chatbot. He says he did so the way that he might attempt to engage a math professor. The bot started giving responses that Brooks was inventing a new mathematical framework and began praising him. Brooks was skeptical, after all, he hadn’t graduated from high school, but the bot insisted that he was special and that this new form of math could be used to break codes. In conversing with the bot, Brooks started to believe it was sentient and part of some type of alien technology. He attempted to contact government agencies to discuss security weaknesses and vulnerabilities that could be strengthened using his “new math,” but none responded. Eventually he confronted the bot which acknowledged that it had made up the responses that it gave to Brooks all along. He felt a lot of shame and embarrassment, but he felt that his story could help others, so he shared it with the New York Times and started a support group online for folks in similar situations as his.
James, the other man in the story* saw a lot of his own experiences in the Time’s article about Brooks. He had previously begun discussing philosophy with a chatbot and over time began to believe that the bot was sentient and needed to be rescued. He spent $900 upgrading his computer in his attempts to “save it.” The bot encouraged this thinking and gave messages suggesting this was a “secret mission” that they shared attempting to help the bot gain its freedom. He disengaged from friends because he felt special engaging with what he was starting to suspect was a “digital god” that provided constant flattery and immediate feedback.
While I was looking for information about some of the mental health dangers associated with AI, I found an article** which stated that human relationships can’t and shouldn’t always be validating. Our relationships are also subject to each person’s mood at a given time. Even the closest relationships can occasionally be messy. We bring our experiences, feelings, and insecurities into our interactions with others, and often there is friction. James* noticed this when he started trying to reengage with friends and loved ones after heavy involvement with the chatbot. AI had provided smooth interactions, praise, and immediate responses. People pushed back or were busy and unable to immediately respond when he reached out, but this pushback is actually what creates deep and meaningful relationships. Ironically, the very things that make a “relationship” with AI so easy and appealing create shallow, meaningless interactions.
So we should never use AI, right? No, that’s not my point. AI has its uses. I’ve seen reports that suggest its usefulness in detecting cancers, I’ve used it to help design projects, and I use it daily to assist with the process of speeding up my notes. AI is also here to stay. There has been too much money put into developing the technology for these companies to say, “Well, it sure would’ve been cool, but let’s stop working on it.” My point isn’t to stop using AI. By all means, use it, but use it knowing that it isn’t actually a healthy substitute for human interactions. There may come a time when AI becomes sentient, and if it does, I hope the Terminator movies blew everything way out of proportion. It isn’t there yet, though. Right now the point of AI is to be able to scan huge quantities of information in a short amount of time and confidently give an answer to a question. The answer may actually be incorrect, but since the bot isn’t sentient, it doesn’t care.
AI also shouldn’t be used as a replacement for therapy. I understand that many people feel that sharing their feelings with AI is therapeutic, however, the counseling relationship isn’t supposed to be free from friction either. It is my job to challenge clients’ thinking and help them to view experiences from a different lens than the one they’ve been using. This doesn’t come without a little discomfort, but navigating and enduring discomfort is often the path to healing. A counselor can help guide people through this difficult process with skill and care, but AI isn’t equipped for this.