OpenAI Warns of Emotional Dependency Risk With New ChatGPT Voice Feature

OpenAI is raising questions about users maybe depending too much on ChatGPT’s new voice tool, which sounds shockingly human-like. Following the recent introduction of the enhanced voice mode to premium users, OpenAI published a safety review emphasizing this concern.

The voice mode is so lifelike that it can tolerate interruptions, react in real-time, even copy conversational sounds like laughter and “hmms.” It can also evaluate feelings depending on the tone of a user. Comparisons to the 2013 film *Her,* where the protagonist has a romantic relationship with an AI, were fast drawn after the movie was revealed.

Now OpenAI worries that such situations might become real instead of fiction. The firm observed that some users of ChatGPT have already started interacting in ways that point to a relationship with artificial intelligence. Although this could provide company for lonely people, it could also sour ties in the actual world.

Although artificial intelligence can still make mistakes, the paper also warns users might rely too much on it merely because it sounds human. This draws attention to a more general problem with artificial intelligence technology: whereas businesses rush to provide fresh AI capabilities, the whole consequences of these developments are yet unknown.

Experts such as Liesel Sharabi of Arizona State University warn that developing close ties with artificial intelligence could prove difficult given the technology is always changing. For its part, OpenAI is dedicated to researching, as it develops its tools, the possibility for emotional reliance on them.