Daily reporter: Wen Qiao Daily editor: Lan Suying On May 20, local time, foreign media NPR exclusively announced that Scarlett Johansson, known as the "widow sister" by movie fans and the AI ​​assistant in the science fiction movie "Her" In a statement released, Scarlett accused

entertainment 4093℃

Every reporter: Wen Qiao Every editor: Lan Suying

On May 20, local time, foreign media npr exclusively announced Scarlett, who is called the "widow sister" by movie fans and plays the AI ​​assistant in the science fiction movie "Her" ·In a statement issued by Johnson, Scarlett accused openai’s chatgpt of illegally using her voice and requested that the voice be removed from the shelves.

Last week, openai released its latest flagship model gpt-4o, which detonated the technology circle. At the press conference, openai demonstrated the latest voice function of chatgpt. Many people found that its expressive synthetic voice sky was surprisingly similar to the voice of the AI ​​assistant played by Scarlett in "Her", which triggered a heated discussion. .

Scarlett stated that in September 2023, when chatgpt launched the voice mode, openai approached her to dub it, but she refused due to personal reasons. OpenAI tried to convince Scarlett just two days before the GPT-4O demonstration, but the new model was released before they could be contacted. Based on this, Scarlett's legal team has previously sent two statements to OpenAI, asking the company to detail its process of developing this sound.

Due to Scarlett’s accusation, openai issued a statement on May 19, local time, announcing that it would suspend the use of this AI sound template. At the same time, in response, the company also announced the detailed creation process of the five dubbing in chatgpt, saying that they were selected from more than 400 professional dubbing and actors.

In fact, this is not the first time openai has faced similar accusations. Earlier, content creators including the New York Times and George R. Martin, the original author of "Game of Thrones", had sued the company in court over copyright issues. As this battle extends to the voice field, foreign media wired believes that this may bring new threats related to deep forgery.

Scarlett: chatgpt "stole" my voice

"When I heard the release demo, I was shocked, angry, and couldn't believe Altman would pursue a project that sounded so similar to me So loud that even my closest friends and the news media couldn't tell," Scarlett wrote in her statement.

Daily reporter: Wen Qiao Daily editor: Lan Suying On May 20, local time, foreign media NPR exclusively announced that Scarlett Johansson, known as the 'widow sister' by movie fans and the AI ​​assistant in the science fiction movie 'Her' In a statement released, Scarlett accused  - Lujuba

Image source: npr reporter bobby allyn's x account

In fact, since openai launched chatgpt's voice mode in September last year, sky's voice has been available. The voice mode of chatgpt has five sounds: breeze, cove, ember, juniper and sky. But it wasn’t until last week when OpenAI demonstrated its latest flagship AI model, the GPT-4O model, which makes the sound more expressive, that the connection between Sky’s voice and Scarlett became clear.

At last week's press conference, openai demonstrated the latest voice function of chatgpt. Thanks to the significant improvement in response speed, this function is almost equivalent to "real-time conversation". Subsequently, Openai CEO Altman released the word "her" on the X platform, implying that the science fiction movie "Her" has become a reality. Many people have realized that the expressive synthetic voice in chatgpt is surprisingly similar to the voice of the AI ​​assistant played by Scarlett Johansson in "Her", which has triggered heated discussions in the outside world.

Image Source:

She said in the statement that in September 2023, when chatgpt launched the voice mode, Altman approached her and asked her if she would be willing to dub chatgpt to comfort users and bridge the discomfort caused by the huge difference between AI and humans. feel. Scarlett said she declined the offer after careful consideration. OpenAI tried to convince Scarlett just two days before the GPT-4O demonstration, but the new model was released before they could be contacted. Based on this, Scarlett's legal team has previously sent two statements to OpenAI, asking the company to detail its process of developing this sound.

openai responded: Suspend sky voice, the 5 voices of chatgpt come from more than 400 submissions

On May 19, local time, openai said that the company is taking measures to solve the problem and is suspending the use of this AI sound template.

The company also announced the detailed creation process of the five dubbing in chatgpt, saying that each voice of has been carefully selected for five months, and is sampled from the cooperating voice actors. , and each voice has been carefully selected for five months. Actors are paid well above the market's highest level.

Daily reporter: Wen Qiao Daily editor: Lan Suying On May 20, local time, foreign media NPR exclusively announced that Scarlett Johansson, known as the 'widow sister' by movie fans and the AI ​​assistant in the science fiction movie 'Her' In a statement released, Scarlett accused  - Lujuba

Picture source: openai official website

openai wrote in the statement: "We believe that AI's voice should not deliberately imitate the unique voice of celebrities - Sky's voice is not an imitation of Scarlett Johansson, but belongs to another professional female Actor, using her own natural speaking voice. To protect their privacy, we cannot share the names of our voice actors.”

It is reported that in May 2023, openai’s casting agency and casting director issued a talent call. In less than a week, they received more than 400 submissions from voice actors and screen actors. Nearly 5 months later, they selected 5 sounds to publish to chatgpt.

The company developed a set of standards for chatgpt’s voices, carefully considering each voice’s unique personality and its appeal to a global audience. These standards for include: actors from diverse backgrounds or who speak multiple languages; a voice that sounds "timeless"; a voice that is approachable and inspires trust; a voice that is rich, warm, and Engaging, confident, charismatic sound; natural and easy to listen to.

Daily reporter: Wen Qiao Daily editor: Lan Suying On May 20, local time, foreign media NPR exclusively announced that Scarlett Johansson, known as the 'widow sister' by movie fans and the AI ​​assistant in the science fiction movie 'Her' In a statement released, Scarlett accused  - Lujuba

Image source: openai official website

deep fake threat or upgrade

"Daily Economic News" reporter noted that Scarlett's accusation is not the first time openai has faced a similar struggle.

In early May this year, several newspapers, including the New York Daily News and the Chicago Tribune, sued OpenAI and Microsoft in New York federal court, accusing them of abusing news content to train generative AI systems.

In December last year, the New York Times filed a copyright infringement lawsuit against OpenAI and its investor Microsoft in the Southern District Court of New York. In the same month, several Pulitzer Prize winners also sued OpenAI and Microsoft for abusing their own works to train large models, pointing out that this The behavior is undoubtedly "scraping" writers' works and other copyrighted materials.

Earlier, in September last year, the Writers Guild of America and 17 famous American writers, including George Martin, the original author of "Game of Thrones", launched a class action lawsuit against openai, claiming that openai used the plaintiff writer without authorization. of copyrighted works to train its AI model. The difficulties faced by

openai also reveal the common problem of the boundary of rights and interests between the entire field of generative AI and content creators.

According to incomplete statistics cited by the media, from November 2022 to October 2023, the U.S. Northern District Court of California alone has accepted 10 cases of copyright holders suing stabilityai, openai, meta, alphabet and other companies for unauthorized use. Cases of model training on copyrighted works.

foreign media wired commented that the conflict between and Scarlett will intensify the existing struggle between openai and artists, writers and other creative people . As this battle spreads into the realm of speech, Wired believes that generative AI makes it easier to create realistic synthetic voices, which may bring new threats related to deepfakes .

In fact, OpenAI said in March this year that it had developed technology that can clone someone's voice from a 15-second clip, but the company said it would not release the technology because it could be abused.

Daily Economic News

Tags: entertainment