Scarlett Johansson says the OpenAI chatbot’s voice is ‘eerily similar’ to her own. By Reuters

(Edited to remove previous reference to OpenAI in paragraph 12 that did not immediately provide an explanation.)

Written by Dawn Chmielewski and Anna Tong

-Scarlett Johansson on Monday accused OpenAI of generating a voice for its ChatGPT system that sounded “eerily similar” to the actress’s after she herself rejected the chatbot’s voice.

Johansson made the comments in a statement released just hours after the artificial intelligence company said it was removing the voice called “Sky.”

OpenAI CEO Sam Altman said in a statement emailed to Reuters on Monday that Sky’s voice was not an imitation of Johansson’s but that of another professional actress.

“Skye’s voice is not Scarlett Johansson’s and was never intended to resemble her voice. We cast a voice actor behind Skye’s voice before reaching out to Johansson,” Altman said.

“Out of respect for Mr. Johansson, we’ve taken a break from using Sky’s voice in our products. We apologize to Mr. Johansson for our failure to communicate better.”

In Hollywood, the fight over the rights to actors’ voices and images is coming into focus as studios consider how to use AI to create new entertainment and computer-generated images and sounds become more indistinguishable from human ones. there is.

In a statement, Johansson said Altman approached her in September and offered to hire her to voice ChatGPT, but she declined.

“After nine months, my friends, family and the general public have all noticed how similar the new system called ‘Sky’ is to mine,” she said.

“When I heard the demo released, I was shocked, outraged and in disbelief that Mr. Altman would go after voices so eerily similar to my best friend and the news media that they couldn’t tell the difference.”

Johansson added that Altman tweeted a reference to the 2013 film “Her,” about a man who develops a relationship with an artificial intelligence assistant, “implying that the similarity was intentional.”

Johansson’s memo was published by journalists from NPR and other news outlets. Her publicist also shared this with Reuters.

She said she hired legal counsel to ask about the process of creating the voice.

OpenAI last week introduced GPT-4o, its latest AI model with audio capabilities that allows users to talk to chatbots and get real-time responses, a significant advance in more realistic AI conversations.