Scarlett Johansson has publicly criticized OpenAI after discovering that the company used a voice in its ChatGPT product that closely resembled hers. Johansson revealed that OpenAI had initially approached her nine months ago to lend her voice to their AI system, but she declined for “personal reasons”. Upon hearing the new voice option, Johansson was “shocked” and “angered” as it was so similar to her own that even her close friends and media outlets couldn’t tell the difference.
Following the backlash, OpenAI removed the voice, named “Sky”, from ChatGPT on Monday. This voice had been prominently featured during the company’s event last week, which showcased the capabilities of the new ChatGPT-4o model. Sky’s engaging and responsive characteristics led users and media to draw comparisons with Johansson’s AI character from the 2013 film “Her”.
Even OpenAI’s CEO, Sam Altman, seemed to acknowledge this resemblance with a cryptic tweet stating “her” after the presentation. However, less than a week later, OpenAI clarified that Sky’s voice was not based on Johansson. The company published a blog post explaining that Sky’s voice belonged to a different professional actress using her own natural voice. OpenAI emphasized its respect for the voice acting industry and the privacy of their voice talents.
Johansson shared that Altman had initially approached her with the idea that her voice could help bridge the gap between technology companies and creatives, making AI more comforting for consumers. Altman even reached out to her agent again two days before Sky’s release, asking her to reconsider. Johansson’s lawyers then contacted OpenAI to have the voice removed.
The similarity between Sky’s voice and Johansson’s did not go unnoticed, with even Johansson’s husband, Colin Jost, making a reference to it on “Saturday Night Live”. The fawning and gendered nature of Sky’s voice also drew criticism, with “The Daily Show” host Desi Lydic humorously commenting on the perceived male-centric design of the AI.
OpenAI defended its choice, stating that the vocals for ChatGPT were selected based on criteria such as having a “timeless” quality and an “approachable” nature that inspires trust. The company had reviewed hundreds of voice acting submissions over five months, eventually selecting five voice options for ChatGPT, which were released in September. The chosen actors were flown to San Francisco for recording sessions to train the AI models.
The removal of Sky’s voice came shortly after several top members of OpenAI’s safety team resigned. One key researcher, Jan Leike, criticized the company for prioritizing “shiny products” over safety. Despite this, Altman and co-founder Greg Brockman defended the company, stating they would not release a product if there were safety concerns.
In its blog post, OpenAI highlighted its collaboration with entertainment industry professionals and the compensation of voice actors. The company, along with others in the AI industry, has faced significant pushback from entertainers, creators, and media companies over copyright violations and fears that AI could replace human jobs. Major unions like Sag-Aftra have even gone on strike over the potential misuse of their likenesses by AI.
This incident underscores the ongoing tension between AI innovation and the rights of artists and creators. OpenAI’s efforts to address these concerns include transparent communication and collaboration with the entertainment industry, though the reaction to the Sky voice highlights the delicate balance between technological advancement and ethical considerations.
