Actress Scarlett Johansson’s voice bears a striking resemblance to OpenAI’s now-pulled “Sky” personal assistant, according to an artificial intelligence lab analysis conducted by researchers at Arizona State University.
At NPR's request, forensic voice experts at the university’s speech lab compared the famous actress’s voice and speech patterns to Sky using AI models developed to evaluate how similar voices are to each other.
The researchers measured Sky, based on audio from demos OpenAI delivered last week, against the voices of around 600 professional actresses. They found that Johansson's voice is more similar to Sky than 98% of the other actresses.
Yet she wasn’t always the top hit in the multiple AI models that scanned the Sky voice.
The researchers found that Sky was also reminiscent of other Hollywood stars, including Anne Hathaway and Keri Russell. The analysis of Sky often rated Hathaway and Russell as being even more similar to the AI than Johansson.
The lab study shows that the voices of Sky and Johansson have undeniable commonalities – something many listeners believed, and that now can be supported by statistical evidence, according to Arizona State University computer scientist Visar Berisha, who led the voice analysis in the school’s College of Health Solutions and the College of Engineering.
“Our analysis shows that the two voices are similar but likely not identical,” Berisha said.
Berisha said while the study analyzed a vast array of subtle vocal features, it also zoomed in on several particular dimensions of each voice and teased out some differences.
The Sky voice has a slightly higher pitch than Johannson’s; the Sky voice tends to be more expressive than Johannson’s voice in the movie Her, and far more expressive than Johannson’s normal speaking voice; and Johannson’s voice is slightly more breathy than Sky’s.
But overall, the two voices have distinct parallels.
The researchers also analyzed the voices for likely “vocal tracts,” or what size throat, mouth and nasal passages would produce a particular sounding voice, and Sky and Johansson had identical tract lengths, they found.
OpenAI and a Johannson spokesman declined to comment on the lab results.
Deliberate design or coincidence?
In the week since Johansson accused OpenAI of copying her voice in its latest version of ChatGPT, a discussion has flared over the influential tech company’s intentions and whether the AI voice in question was the product of deliberate design or a coincidence.
As the debate raged, OpenAI paused its Sky voice, which was one of five possible voice options as part of the newest ChatGPT version.
OpenAI maintains that Sky was not created with Johansson in mind, saying it was never meant to mimic the famous actress.
“It's not her voice. It's not supposed to be. I'm sorry for the confusion. Clearly you think it is,” Altman said at a conference this week. He said whether one voice is really similar to another will always be the subject of debate.
Johansson, for one, is outraged by the similarity. She said that Altman twice reached out to her attempting to license her voice for ChatGPT. She decided against it.
When the company’s updated personal assistants were unveiled, she said she was shocked over how “eerily similar” the Sky voice resembled her own.
Altman himself amped up the speculation on the day of the release by posting on X one word, “her,” the name of the 2013 romantic sci-fi film in which a lonely man falls in love with a superintelligent computer operating system voiced by Johansson.
OpenAI says it hired a voice actress to help develop the Sky voice months before Altman began reaching out to Johannson, which was first reported by the Washington Post.
OpenAI would not publicly identify the actress, citing personal privacy and potential risks to her safety.
Altman, last year, said Her, featuring Johansson, was his favorite movie about AI, complimenting the film for being “prophetic” about how conversational AI chatbots would one day function in people's lives.
Still unknown is whether Altman ever privately told the team that led the casting for the Sky voice that Johansson should be an inspiration.
OpenAI says Altman was not closely following the casting process. The company says he deferred to OpenAI Chief Technology Officer Mira Murati, who told NPR she didn’t even know what Johannson sounded like until people were comparing Sky to the actress.
But Johansson said when Altman engaged her about potentially having her voice licensed, he pitched it as a way to make the cutting-edge conversational bot “comforting to people” who are perhaps wary of AI services that behave like humans — Altman’s last outreach came just two days before the public launch of the new personal assistant, Johansson said.
"After much consideration and for personal reasons, I declined the offer," she said.
Copyright 2024 NPR