Financial News Agency, May 21 (Editor Shi Zhengcheng) On Monday local time, the American technology innovation company OpenAI issued a statement announcing that because the female dubbing of its chatbot ChatGPT named Sky has caused widespread doubts among netizens, the company is

Financial News Agency, May 21 (Editor Shi Zhengcheng) On Monday local time, the American science and technology company openai issued a statement announcing that the company is taking measures because the female dubbing of its chatbot chatgpt named sky has caused widespread doubts among netizens. To solve the problem, the use of this ai sound template is currently being suspended.

(Source: x)

It should be noted that the voice mode of chatgpt was actually launched at the end of September last year. A total of 5 dubbing models were released, and the controversial sky was also among them. Of course, the source of the controversy was the “real-time voice dialogue” introduced by OpenAI at the press conference last week. In addition to the similar sound,

is also suspected of being deliberately "flattering"

. At last week's press conference, openai released the latest gpt-4o model. By doubling the response speed, the voice function of chatgpt is as smooth as 's "real-time conversation" - you can interrupt the chatbot from time to time, as if you are communicating with a real person. Moreover, chatgpt can also distinguish the speaker's tone and respond with stronger emotions.

(Researchers demonstrate voice dialogue live, source: openai spring update)

Perhaps to celebrate the leap-forward progress in human-computer interaction, openai CEO Sam Altman immediately sent out a word after the conference - —her, referring to the scene from the science fiction romance movie "her" that is becoming a reality. That 2013 film tells the story of a human and an artificially intelligent assistant who fall into a complicated love affair.

(Source: "The artificial intelligence assistant voiced by Scarlett Johansson is very similar to .

(Scarlett Johansson data picture, source: social media)

In addition to sound, openai’s new conversation function has also caused a deep level of controversy: in addition to interacting with the host, chatgpt seems to show additional “pleasure” the interlocutor's tendencies. For example, when demonstrating the multi-modal conversation function, the chatbot will exclaim "Wow, the clothes you are wearing are really stylish." When receiving compliments, chatgpt will also say "Don't be like this, you made me blush."

Some netizens pointedly pointed out: This makes people feel like a female character written for men .

Due to the space limit of the press conference, and the new voice mode has not yet been pushed to paying users, it is not yet known whether chatgpt will have a side to please female users in voice mode.

Regarding these controversies, openai also published a long article on Sunday. Although it did not respond to the question of "flattering" in the dialogue, it clearly and firmly denied that the voice sounded like an actress.

openai reveals: Where do these voices come from

In the article "How are the voices of chatgpt selected?", openai disclosed that before launching the voice mode in September last year, the company conducted a 5-month casting process, and finally selected from 400 Five voices were selected from multiple applications.

openai categorically stated that the voice of artificial intelligence should not deliberately imitate the unique voice of celebrities - sky's voice is not imitating Scarlett Johansson, but belongs to another professional actress, which is her natural speaking voice. openai also stated that in order to protect the privacy of voice actors, the company cannot disclose their names to the public.

openai revealed that in early 2023, in order to let chatgpt "speak", the company worked with a series of well-known casting directors and producers to set standards for selecting "chatgpt voice actors". For example:

has different backgrounds or can speak multiple languages. An actor of language;

A voice that sounds "timeless";

A voice that is approachable and inspires trust;

A voice that is rich in timbre, warm, engaging, confident, and charming;

is natural and easy to listen to.

openai said that in May 2023, the casting agency issued a notice to recruit voice actors. In less than a week, a total of more than 400 applications were received, including from professional voice actors and some film and television actors.

To audition, actors were asked to record a script of chatgpt responses, covering issues including mindfulness, brainstorming travel plans, and conversational text related to the user's "normal day."

Through the audition, openai determined a preliminary list of 14 people. Next, the company discussed AI voice interaction and the company's vision with each actor, including technical capabilities, limitations, risks involved, and safeguards that have been implemented. Afterwards, openai’s internal team finally selected 5 voices based on product and research perspectives. The actors flew to San Francisco to record in June and July last year.

openai also emphasized in the announcement that each voice actor in has been paid "above the highest level in the market", and this treatment will continue as long as chatgpt continues to use their voices.

The company said it plans to provide paying customers with access to gpt-4o's new voice mode in the coming weeks. In the future, more voices will be introduced to chatgpt to better match the different interests and preferences of users.