We’re shocked.
A+ Copycat
OpenAI’s controversial “Sky” voice for ChatGPT sounds remarkably similar to the voice of Scarlett Johansson, a forensic analysis has found, adding weight to what many already suspected and what Johansson herself has charged: that OpenAI deliberately mimicked the actress’s voice without her permission.
The analysis, conducted by researchers at the Arizona State University and commissioned by NPR, used several AI models to evaluate similarities between the voices of Sky and about 600 actresses, including Johansson.
Lo and behold, it found that Johansson’s voice was more similar to Sky than 98 percent of the other candidates.
There are a few caveats, however. Johansson wasn’t always the top scorer, with the voices of Anne Hathaway and Keri Russell “often” being rated as more alike, according to NPR. Sky’s voice is slightly higher pitched and more expressive, too, while Johansson’s is breathier.
But other parts of the analysis are damning, such as one that simulated the speakers’ vocal tracts based on the characteristics of their voice and found that Sky and Johansson would have identical tract lengths.
Visar Berisha, a computer scientist at ASU who led the analysis, summed it up neatly. “Our analysis shows that the two voices are similar but likely not identical,” he told NPR.
Sky-High Lies
The controversy stems from a big update to ChatGPT released last month, which debuted a new voice assistant capable of real-time conversation.
Sky was one of those voices, and soon enough, people took note of its resemblance to Johansson’s role in the sci-fi movie “Her,” in which she voices a chirpy AI chatbot that the film’s melancholic protagonist falls in love with.
If those parallels weren’t already suspicious, they were all but confirmed by OpenAI CEO Sam Altman — a professed fan of the movie — who cheekily tweeted the single word “her” on the day of the voice assistant’s release.
Then in a blundering series of backpedals, the AI company suddenly pulled the Sky voice days later, but said it had not copied ScarJo’s voice. Instead, it claimed, a different actress was behind the chatbot (which was later corroborated by reporting from The Washington Post).
Johansson fired back, revealing that OpenAI had in fact twice approached her to license her voice. She turned the offers down, only to discover that OpenAI had released a chatbot with a voice she thought was “eerily similar” to hers.
In the face of mounting negative PR, OpenAI has maintained that this whole fiasco was simply the fault of its poor communication with the actress. Johansson hasn’t filed a lawsuit yet, but she has hired lawyers. Many legal experts already believed that she would have a strong case. And now, with these latest forensic findings, it could be even stronger.
More on AI: OpenAI Insiders Say They’re Being Silenced About Danger
Share This Article