“We’ve been in conversations with ScarJo’s team because there seems to be some confusion. We want to take the feedback seriously and hear out the concerns,” says Joanne Jang, model behavior lead at OpenAI.
OpenAI pulled Sky, one of the default voices on ChatGPT, after it made headlines last week for sounding just like Johansson in her role as an intelligent voice assistant in the movie Her. OpenAI CTO Mira Murati told The Verge in an interview last week that the company did not intend to mimic Johansson’s sultry AI assistant. The Verge has reached out to Johansson’s team for comment.
Jang, who is part of the team that casts actors to provide voices to the chatbot, says that people may be hearing similarities between Johansson’s character and OpenAI’s voice because there are few examples of convincing female voice assistants around. Other voice assistants, like Siri and Alexa, “still sound robotic,” Jang says.
ChatGPT has had voices since last year, but Sky’s voice did not gain prominence until last week when the company made voice assistants more expressive through its GPT-4o model. Jang says users can prompt their chosen voice to have a personality, so Sky or any of the other voices on ChatGPT can be peppy, more measured, or even sad, depending on the user and the personality they most connect with.
“We want users to feel at ease,” Jang says, and customization is important to making the experience feel natural to users. “We know they want a product that [doesn’t make users] have to pretend to be someone else.”