NEW YORK (AP) – OpenAI announced Monday that it plans to stop using one of the ChatGPT voices that “Her” actor Scarlett Johansson claims is “eerily similar to her own.”
in post OpenAI said it is working on “suspending” Sky, the name of one of five voices ChatGPT users can choose to speak, on social media platform X. The company said it had “heard questions” about its flagship artificial intelligence chatbot, particularly how it selects the realistic voice options available on Sky, and wanted to address them.
Among those raising questions was Johansson, famous for voicing a fictional, then-futuristic AI assistant in the 2013 film. “she.”
Johansson issued a statement saying that OpenAI CEO Sam Altman approached her in September and said it would “provide reassurance” to those who have concerns about the technology, adding that the system would have a voice. He said that he had asked if he could lend it to him. She said he declined her offer.
“When I heard the public demo, I was shocked and angry, and I couldn’t believe that Mr. Altman would go after a voice so eerily similar to mine that close friends and the press couldn’t tell it apart.” “It was,” Johansson said.
OpenAI “reluctantly” agreed to remove Sky’s audio, Altman said, after hiring a lawyer to write to Altman and asking him how the company came up with Sky’s audio. Stated.
OpenAI had moved to debunk the internet’s theories about Johansson in a blog post accompanying an earlier announcement that aimed to detail how ChatGPT’s audio was chosen.company I have written The company believed that AI voices “should not deliberately imitate a celebrity’s distinctive voice,” and claimed that Sky’s voice belonged to “another professional actress.” However, it added that the name of the expert could not be published for privacy reasons.
In a statement sent to The Associated Press following Johansson’s response late Monday, Altman said OpenAI cast the voice actors for Sky “prior to any outreach” to Johansson.
“Skye’s voice is not Scarlett Johansson’s voice, and it was never intended to sound like her,” Altman said. “Out of respect for Ms. Johansson, we have suspended the use of her Sky voice in our products. We apologize to Ms. Johansson for not communicating with her better.”
San Francisco-based OpenAI first released voice capabilities for ChatGPT in September, including five different voices, allowing users to have two-way conversations with an AI assistant. “Voice mode” was initially available only to paid members, but in November OpenAI Announced The feature will be free for all mobile app users.
And ChatGPT interactions are becoming increasingly sophisticated. Last week, OpenAI said: Latest updates to generative AI models It can imitate human rhythms in verbal responses and can even try to detect people’s moods.
OpenAI says its latest model, called GPT-4o, runs faster than previous versions and can infer text, audio, and video in real time. In a demonstration during OpenAI’s May 13 announcement, the AI bot chatted in real time and added emotion, specifically “more drama,” to its voice on request. They also tried to infer a person’s emotional state by looking at a selfie video of their face, while helping with things like language translation and step-by-step math problems.
GPT-4o, short for “omni,” is not yet widely available. We plan to select users in stages over the coming weeks and months. The text and image features of this model have already started rolling out and will also be available to some users of ChatGPT’s free tier, but the new voice mode will only be available to paid subscribers of his ChatGPT Plus. will be available.
Most people haven’t gotten their hands on these newly announced features yet, but they’re coming from Spike Jonze’s film about an introvert (Joaquin Phoenix) who falls in love with a human who controls an AI. It has further evoked comparisons to the dystopian romance “Her”. system (Johansson) causes many complications.
Altman also seems to take advantage of this – simply Post On the day of GPT-4o’s announcement, the word “she” appeared on social media platforms.
Many people who responded to the model’s demo last week also noticed that some of the interactions had an oddly flirtatious feel to them. in one video A female-voiced ChatGPT posted by OpenAI, for example, praises an employee for “wearing an OpenAI hoodie,” while another chatbot To tell When someone says something amazing to me, I say, “Stop, you’re blushing.”
The incident sparked a debate about the gender-biased ways technology companies have long developed and used voice assistants. Its origins go back long before the latest wave of generative AI advanced the capabilities of his AI chatbots. Cultural and Scientific Organization of the United Nations in 2019 Pointing out “fixed obedience” Even in the face of sexist insults and harassment, it’s built into the default female voice assistants (like Apple’s Siri and Amazon’s Alexa).
“This is clearly programmed to feed the egos of these guys,” Daily Show senior correspondent Desi Riddich said of GPT-4o in a segment last week. “It really shows that humans are the ones who developed this technology.”
