image credit:MLYearning
OpenAI surprised users this year by showing an upgraded speech option for ChatGPT’s most advanced version.
The ChatGPT advanced voice mode sounds very human, unlike Alexa or Siri. It responds in real-time, adjusts to interruptions, generates giggling noises during jokes, and can assess emotional condition by tone of voice. (The initial demo also sounded suspiciously like Scarlett Johansson).
Paid customers will get advanced voice mode for ChatGPT-4o, the most powerful chatbot, on Tuesday. Advanced speech mode will be introduced to a small number of app “Plus” customers in the fall, with the goal of making it available to all Plus users.
ChatGPT already has a simpler voice mode. The launch of a more advanced voice mode could transform OpenAI from a major AI chatbot into a virtual, personal assistant that users can talk to like friends. ChatGPT’s advanced speech mode may stimulate more use and rival Apple and Amazon’s virtual assistants.
Adding a sophisticated voice capability to ChatGPT raises huge questions: Will it understand users’ speech despite speech differences? Will customers blindly trust a human-sounding AI assistant even when it makes mistakes?
OpenAI had planned to debut the sophisticated speech mode in June, but it required “one more month to reach our bar to launch” to evaluate the tool’s safety and assure it can be used by millions of people while retaining real-time responses.
The business claimed Tuesday that more than 100 testers, “who collectively speak a total of 45 different languages and represent 29 different geographies,” have tested the AI model’s vocal capabilities in recent months to find shortcomings.
To prevent impersonation, the business said voice mode won’t be allowed to utilize voices beyond four pre-set options established with voice actors and will restrict requests to recreate music or other copyrighted material. The application will have the same restrictions as ChatGPT’s text mode to avoid illegal or “harmful” information, according to OpenAI.
Advanced speech mode will also lack the voice that many, including the actor, thought sounded like Johansson, unlike OpenAI’s demo in May. After the actor complained, OpenAI halted use of the voice “out of respect,” even though it was never meant to sound like Johansson.
ChatGPT’s advanced speech mode follows OpenAI’s announcement last week that it was piloting a search engine using its AI technology as it expands its consumer-facing AI products. The OpenAI search engine could challenge Google’s search dominance.
Comment Template