Scarlett Johansson has voiced her disapproval of OpenAI for using a voice in its ChatGPT product that closely mimics her own. The 'Avengers' and 'Jojo Rabbit' actor revealed that OpenAI had approached her nine months ago to lend her voice to its AI system, an offer she declined for personal reasons. However, she was "shocked" and "angered" to discover that the AI's voice option, named "Sky," sounded remarkably similar to hers, causing confusion among her friends and media outlets.
How Did OpenAI React?
OpenAI has since withdrawn the Sky voice from ChatGPT following widespread criticism. The voice had been prominently featured in a recent event showcasing the new ChatGPT-4o model, where it was noted for its personable and responsive nature.
Did OpenAI Really Use 'Her' Voice?
Comparisons were quickly drawn between Sky and Johansson’s AI character in the 2013 film "Her."
Sam Altman, CEO of OpenAI, seemed to hint at the intentional resemblance by tweeting "Her" after the presentation.
Following the backlash, OpenAI issued a clarification stating that Sky’s voice was not based on Johansson. According to a blog post, the voice was provided by a different professional actress whose identity remains undisclosed for privacy reasons.
Johansson recounted that Altman had initially suggested her involvement could bridge the gap between technology and the creative community, potentially easing public apprehension about AI. Despite a last-minute attempt by Altman to persuade her to reconsider, Johansson's legal team intervened to ensure the removal of the voice.
ChatGPT 'Flirting' With Users?
The Sky voice also faced scrutiny for its flirtatious and gendered tones, leading to broader discussions about the design choices in AI voices. Critics, including "The Daily Show" host Desi Lydic, highlighted the gendered nature of the voice, questioning the implications of such design decisions.
OpenAI defended its selection process, explaining that Sky's voice was chosen for its "timeless" and "trustworthy" qualities. The company reviewed hundreds of voice submissions over five months, ultimately choosing five voices for ChatGPT, with recording sessions conducted in San Francisco.
Sudden Resignations
The controversy surrounding Sky’s voice emerged just as several key members of OpenAI's safety team resigned. Jan Leike, a notable researcher, criticised the company for prioritising product appeal over safety. In response, Altman and co-founder Greg Brockman asserted that OpenAI would not release products if there were safety concerns.
In its Sunday blog post, OpenAI emphasised its collaboration with the entertainment industry and the compensation of voice actors, amidst growing legal and ethical challenges faced by AI companies. These challenges include lawsuits and strikes from major entertainment unions like SAG-AFTRA, concerned about the impact of AI on human labour and likeness rights.