Scarlett Johansson Challenges OpenAI Over Voice Similarity to Hers
American actress Scarlett Johansson has raised concerns over the uncanny resemblance of OpenAI's latest voice assistant, dubbed “Sky”, to her own voice. OpenAI rolled out the latest version of ChatGPT, known as GPT-4o, which features advanced performance and human-like behavior, igniting debates over its potential ethical implications. Johansson's lawyers have reached out to OpenAI, demanding clarity on how Sky's voice was created, particularly after numerous observers noted striking similarities to the actress' voice from the 2013 film Her.
The actress, through her agent, emphasized that she declined an offer from Sam Altman, head of OpenAI, to lend her voice to ChatGPT. Despite her refusal, Johansson was shocked and angry upon hearing the demonstration of Sky, where even her close friends couldn’t distinguish the voice from hers. In response, Johansson accused Altman of pushing forward with a voice effort that seemingly mimics her intonation, leading to substantial public confusion.
OpenAI Responds to Allegations
In a formal statement, OpenAI has denied intentionally imitating Johansson's voice. The company stated that Sky's voice was synthesized using inputs from various professional voice actresses, not specifically Johansson's. To address the controversy, OpenAI has announced that they are temporarily suspending Sky's voice to review and amend the voice selection process. Their blog post asserts, “We believe that AI voices should not intentionally imitate those of celebrities,” clarifying that Sky's voice is a unique creation by a professional actress whose privacy they wish to protect.
Amid the public outcry and ethical concerns surrounding deepfakes, Johansson has called for greater transparency and legal protections to safeguard individuals' identities and work from such digital replications. She has demanded the implementation of laws that would protect individual rights against unauthorized use of their likeness in AI technologies.
The recent developments at OpenAI come alongside their decision to dissolve a team focused on mitigating the long-term risks of advanced AI. This move, which resulted in the departure of significant figures like co-founder Ilya Sutskever and team head Jan Leike, underscores the urgency and complexities entwined in the realms of AI safety and ethical use.
- OpenAI's statement shed light on the meticulous process they followed to craft ChatGPT's voices. Beginning in early 2023, they sought actors with diverse backgrounds and multilingual abilities. A central criterion included having a voice that is timeless, warm, and charismatic. Out of 400 proposals, the five chosen voices were recorded for ChatGPT in San Francisco last summer.
- Though Sky's voice has been temporarily suspended, OpenAI plans to introduce additional voice options in future releases to cater to varied user preferences. They have reassured that all voice actors receive compensation above market rates and will continue to benefit as long as their voices are utilized in OpenAI's products.
