OpenAI said it plans to pull a speaking voice eerily similar to actress Scarlett Johansson from its ChatGPT service.
The maker of the popular AI-powered bot has decided to suspend the use of Sky Voice in the audible version of ChatGPT after users compared it to Johansson’s performance as a fictional virtual assistant in the 2013 film Her. announced that it did. This is the story of a man who falls in love with an AI system.
OpenAI added Sky and four other voices as part of an updated version of GPT-4o. This allows users to ask questions orally and get responses by voice.
“We believe that AI voices should not intentionally imitate the distinctive voices of celebrities.” The company wrote in a blog post: The move was announced over the weekend.
OpenAI said Skye’s voice is “not an imitation of Scarlett Johansson” and “another professional actress using her own natural speaking voice.”
The company said last year that it worked with casting directors and producers to find voice actors who met certain criteria, including those with “diverse backgrounds.”
OpenAI said it was looking for a “timeless feel” and a “relatable” voice to “inspire trust.”
The casting call received 400 applications, which were ultimately narrowed down to 14 actors. For the next stage, OpenAI selected her five to represent the voices of Breeze, Cove, Ember, Juniper, and Sky.
The company then flew five actors to San Francisco and began a five-month process that included recording sessions and in-person meetings, according to a blog post.
New audio features in GPT-4o allow users to talk to ChatGPT and get real-time responses without delay, as well as interrupt ChatGPT during a conversation. This is a characteristic of realistic conversations that AI voice assistants have found difficult.
“I feel like an AI in a movie… Talking to computers has never come naturally to me. Now it’s here,” OpenAI CEO Sam Altman blogs. I wrote in the post.
Microsoft-backed OpenAI faces increased competition and pressure to grow the user base of ChatGPT, a popular chatbot product that has wowed the world with its ability to generate human-like written content and top-notch software code. confronting.
During a livestream event last week, OpenAI researchers showed off new voice assistant features for ChatGPT.
In one demo, ChatGPT used visual and audio capabilities to converse with researchers as they solved mathematical equations on paper.
In another demonstration, the researchers demonstrated the real-time language translation capabilities of the GPT-4o model.
The OpenAI demonstration verged on science fiction, with ChatGPT and his interlocutor engaging in coquettish banter at one point.
OpenAI researchers told the chatbot they were in a great mood because it demonstrated “how helpful and awesome you are.”
ChatGPT replied: You’re making me blush! “
Mira Murati, OpenAI’s chief technology officer, said at the event that the new model is more cost-effective than the company’s previous models, so it will be made available for free.
GPT-4o’s paid users have larger capacity limits than the company’s free users, he said. The GPT-4o model will be available on ChatGPT in the coming weeks, the company said.
Additionally, free ChatGPT users now have access to a “Browse” feature that allows ChatGPT to display the latest information from the web, Murati told Reuters after the event.
Murati said the company does not intend to profit from free users through advertising sales.
Shortly after its release in late 2022, ChatGPT was said to be the fastest application in history to reach 100 million monthly active users.
But global traffic to ChatGPT websites has been on a roller coaster over the past year, and is finally returning to its May 2023 peak, according to analytics firm Similarweb.
Last week, Google revealed how it is building artificial intelligence across its businesses, including a beefed-up Gemini chatbot and improvements to its vaunted search engine to compete with AI rivals.
Google’s latest salvo includes adding to its family of Gemini 1.5 AI models, known as Flash, which are faster and cheaper to run. The prototype, called Project Astra, can talk to users in real time about anything captured by their smartphone’s camera. Search results are categorized under AI-generated headings.
with post wire
