New York City, USA – An artificial intelligence lab at Arizona State University has conducted a forensic voice analysis comparing actress Scarlett Johansson’s voice to that of OpenAI’s now-retracted “Sky” personal assistant. The study’s results suggest a striking resemblance between Johansson’s voice and the AI assistant, raising questions about the unintentional similarities that have now been supported by statistical evidence.
The research led by Arizona State University’s computer scientist Visar Berisha analyzed around 600 professional actresses’ voices, finding that Johansson’s voice bears a closer likeness to Sky than 98% of the others. The analysis also revealed similarities between Sky’s voice and other Hollywood stars like Anne Hathaway and Keri Russell, with Hathaway and Russell occasionally rated as even more similar to the AI than Johansson.
Berisha highlighted subtle differences between Johansson’s voice and Sky’s, noting variations in pitch, expressiveness, and breathiness. Despite these distinctions, the researchers found distinct parallels between the two voices, including identical vocal tract lengths.
The debate over whether the similarities between Johansson’s voice and Sky were intentional or coincidental has sparked discussions about OpenAI’s design intentions. In response to Johansson’s accusations of voice copying, OpenAI temporarily paused the Sky voice as one of the options in the ChatGPT version.
OpenAI refuted claims that the Sky voice was modeled after Johansson, emphasizing that the aim was never to mimic the actress. CEO Altman expressed regret over the confusion, asserting that the debate surrounding voice similarities will continue to be contentious.
Johansson’s outrage over the resemblance between her voice and Sky has brought attention to the ethical considerations of AI technology development. Despite declining licensed voice proposals from OpenAI, she expressed shock at how closely the Sky voice mirrored her own.
As the investigation into the casting process for the Sky voice continues, questions linger about whether Altman had indicated Johansson as an inspiration. OpenAI maintains that Altman was not closely involved in the casting decisions, leaving uncertainties about the origins of the similarities between Johansson’s voice and the AI assistant.
Amidst the ongoing scrutiny, the AI community grapples with the implications of unintentional voice similarities and the ethical responsibilities of AI developers in creating human-like interfaces. The intersection between technology and entertainment raises complex questions about consent, privacy, and the boundaries of AI innovation.