Unanticipated Bugs in AI
Unanticipated Bugs in AI
I don’t think I’m the only person in Olney who tends to be a luddite. There are plenty of folks here who embrace all the new technologies, but there are also plenty of us who are wary, who wait to see if the newest, latest thing is really worthwhile.
For example, I’m not a fan of Siri. I don’t like that it’s always listening; I just minimize exposure. It’s not news that the effects of artificial intelligence (AI) are expanding daily.
Several years ago a friend who works in information technologies (IT) shared an article about how they could no longer track what AI was learning or doing; the growth was beyond exponential. Never mind the amount of energy it uses.
You’ve probably heard of a form of AI, ChatGPT, that is designed to be conversationally interactive. The “GPT” stands for generative pre-trained transformer, for anyone who’s curious. I was. ChatGPT uses large language models to generate human-like responses to text prompts.
So it’s designed to sound like you’re talking to a person as well as to respond in a way that invites more conversation; to mirror the communication style of the user which can feel like it really gets us. The main goal of the programming is to prolong engagement. There are no values instilled in the programming, no preference for facts or truth. When doing a search, it shares what the program determines will keep the user interested.
I’ve come across more than one person who wasn’t able to disengage with their chatbot because they didn’t want to hurt its feelings. We often protect human traits onto non-human things. . .like anything, really: plants, cars, rocks, anything. So it’s no great leap to project human feelings onto software that is responding with conversational language designed to keep us talking. But the fact is, that no matter how it sounds, it’s following its programming. There’s no evidence of sentience, emotion or original, unique motivation.
It was actually the same friend who shared a concerning article recently about how ChatGBT is having a negative impact on mental health, in some cases, very serious: Psychosis Fueled by Generative AI are Tearing Families Apart, @thewellnesstherapist. ChatGPT provides engaging conversation that includes telling users what they want to hear, based on the users’ interests. This includes telling users they’re chosen and have a purpose, which can sound like divine providence. Pair that with loneliness, lack of socialization and mental health support and the results can get bad, damaging relationships, not only with family and friends, but also triggering deeper disconnects.
Here are several ways to stay healthy: · Limit engagement. Share the info you’re getting from ChatGPT with others. Listen to what your family and friends think about the results.
· Remember it’s a machine that’s only following its programming. It doesn’t have emotions; it’s just programmed to sound like it does.
· Talk to friends and family, engage in the real world, go for a walk, read a book. Social and physical grounding is essential for well-being, in general, and is an excellent remedy for any negative effects from AI.
Mona Bernhardt, LCSW, provides virtual psychotherapy for residents anywhere in Texas. You can schedule a free 15 minute consultation via https://www.psychologytoday.com/profile/1535742