OpenAI has announced the rollout of its highly anticipated advanced voice mode for ChatGPT, giving ChatGPT Plus subscribers access to enhanced voice capabilities. This feature, which was showcased at OpenAI's GPT-4o launch event earlier this year, has undergone improvements and safety refinements before its release. In this blog post, we will explore the key highlights of this new advanced voice mode and the additional measures that OpenAI has implemented to ensure user safety.
The New and Improved Voice Mode
During OpenAI's GPT-4o launch event, the advanced voice mode demonstrated a significant improvement over ChatGPT's previous voice capabilities. OpenAI employees showcased the chatbot's ability to adapt to interruptions and respond accordingly, making conversations with the AI feel more dynamic and interactive.
Delay and Safety Concerns
Although the advanced voice mode was initially planned to release in alpha in late June, OpenAI decided to postpone the rollout by one month. This delay allowed the company to further refine the voice model and strengthen its ability to identify and refuse certain content. OpenAI acknowledges the importance of ensuring the safety of its technology, particularly in light of recent scrutiny and concerns.
Red Teaming and Safety Measures
To address safety concerns, OpenAI engaged over 100 external red teamers, individuals who specialize in detecting vulnerabilities in technologies, to test the capabilities of the voice model. This rigorous testing process aimed to identify any potential weaknesses and ensure that the voice mode aligns with OpenAI's safety standards.
As part of its commitment to protecting against potential misuse, OpenAI has implemented new filters to recognize and block requests that involve generating copyrighted audio or music. These additional measures further enhance the safety and compliance of the advanced voice mode.
Addressing the Scarlett Johansson Criticism
During the GPT-4o launch event, one of the criticisms surrounding the advanced voice mode was its similarity to the voice of Scarlett Johansson in the movie "Her." Responding to these concerns, OpenAI spokesperson Taya Christianson clarifies that ChatGPT's new mode will only utilize four preset voices created with voice actors. This decision ensures that ChatGPT cannot impersonate the voices of individuals or public figures. Any outputs generated by the AI that differ from these preset voices will be blocked.
Future Plans
OpenAI's goal is to make the advanced voice mode available to all ChatGPT Plus users in the fall. By rolling out this feature gradually and addressing safety concerns, OpenAI aims to provide an improved and secure user experience.
Summary
OpenAI's launch of the advanced voice mode for ChatGPT Plus marks an important milestone in the development of AI-powered conversational assistants. The refined voice capabilities, paired with enhanced safety measures, ensure a more engaging and secure experience for users. As OpenAI plans to expand the availability of this feature to all ChatGPT Plus subscribers, we can expect AI conversations to become even more seamless and lifelike in the near future.
