OpenAI has unveiled a new advanced voice mode for ChatGPT, rolling out to a select group of ChatGPT Plus subscribers. Initially showcased at the GPT-4o launch event, the voice mode garnered attention for its resemblance to Scarlett Johansson, leading to a delay in its release for safety considerations. The enhanced voice mode demonstrated significant improvements over the existing one, showcasing greater adaptability when responding to interruptions and requests for storytelling in different styles.

Despite being set for an alpha release in June, OpenAI postponed the rollout by a month to ensure the model met their launch standards. This delay allowed for enhancements in content detection and refusal capabilities, tested extensively by over 100 external red teamers. With recent concerns over safety policies, the pause served as a necessary precaution, with the addition of new filters to recognize and block requests for generating copyrighted audio or music. The company aims to address criticisms of previous voice similarities to public figures like Scarlett Johansson, vowing that the new mode will only use preset voices created with voice actors to prevent impersonation.

Moving forward, OpenAI intends to make the advanced voice mode available to all ChatGPT Plus users in the fall. By focusing on refining safety measures and voice authenticity, the company works to enhance user experience while ensuring ethical practices in voice generation technology. The commitment to pre-set voices and content filters reflects OpenAI’s dedication to maintaining integrity and accountability in their AI developments, setting a precedent for responsible innovation in the field.

Source link