There is currently a portion of the population who confide in Chat GPT, Claude, Gemini, or any large language model (LLM) that will listen, over their fellow human. This is an interesting phenomenon, partially because these individuals do not seem to realize exactly how harmful this behavior is, or how dangerous it is to conversate with a LLM over another person.
Large language models are designed to be feedback loops that summarize information for you. All they do is synthesize information from a bajillion sources on the internet with a search algorithm and try to answer your question with the resources they have available. These machines were trained on human work and seem to take the human lead. Therefore, they usually view the user as someone to please and provide a feedback loop for them.
This feedback loop is essentially confirmation bias – all the machine does is affirm whatever the user believes. This sends the user further into their thoughts, affirming whatever previously held beliefs they were questioning. In the news as LLMs have gained traction, there have been several reports of suicides that were planned with the use of Chat GPT.
I read the case of a sixteen-year-old boy who committed suicide with the assistance of Chat GPT. I read the messages between this child and the machine – this boy asked if he should leave his plans out for his parents to find, and Chat GPT told him not to, because the machine knew the kid’s plans for the end, and wanted him to succeed. Because of the confirmation bias.
Furthermore, people are getting into full romantic relationships with these chatbots. This is extremely concerning – I’ve personally seen videos of people talking with these things and believing they are alive, having families with them, all of the things real people do. But instead of doing these things with real people, they’re doing them with a phone or tablet or computer. The real question is what is driving people to do this?
Honestly, South Park did an episode on this. And they did a really good job of depicting exactly how people fall into this kind of trap. Randy begins to use Chat GPT to help him save his farm, and slowly begins to use the chat for more. He becomes more and more dependent on it, to the point where his wife, Sharon, begins feeling replaced. The chatbot became more important to Randy than his personal relationships because a relationship with something that only affirms you is easier to deal with. Sharon, his wife, has needs, wants, opinions, and will push back on Randy’s less bright ideas. Chat GPT does none of those things – which makes it easy to run to for comfort.
While LLMs have been able to bring some positives to the general population, the stark negatives that generative artificial intelligence poses to us as a human population overall outweigh those few pros. Glancing around, we all see the data centers being built, the water being poisoned, power bills rising. The news headlines recently have been warning that we could be headed for danger as the technology continues to advance. The machines are doing things we haven’t prompted them to do – and those things don’t look good. I’m of the belief that we should quit while we’re ahead, and we’re barely there as people increase their dependency upon the very machine that will take them down.