OpenAI's new voice model wants you to talk over it

In a groundbreaking livestream event, OpenAI introduced GPT-Live, a cutting-edge voice model enhancing the ChatGPT Voice experience. This innovative update aims to make interactions with AI feel more akin to conversing with a real person. Unlike traditional systems that wait for the speaker to finish before responding, GPT-Live can listen and respond simultaneously, creating a more fluid dialogue. The model enriches conversations by incorporating real-time acknowledgments such as 'mmhm,' 'yeah,' and 'got it' while the user is still speaking. During a demonstration, a user prompted ChatGPT to verify the date of an upcoming meeting while simultaneously checking weather and traffic conditions. Remarkably, ChatGPT maintained the conversation flow by providing brief affirmations like 'hmm' and 'sure' as new requests were introduced. Additionally, GPT-Live excels in real-time language translation, a task previously constrained to turn-based responses. This advancement allows for a more dynamic and engaging interaction, with the AI keeping pace as discussions evolve. OpenAI's president, Greg Brockman, referred to this development as a 'much more natural way of interacting with your computer.' OpenAI is not alone in this pursuit; other companies, such as Thinking Machines, are also working on similar advancements. Led by former OpenAI chief technology officer Mira Murati, Thinking Machines is developing interaction models that facilitate continuous input and output across various media, moving away from the rigid structure of traditional chatbots. As AI language models increasingly become subjects of humor online—often criticized for their overly enthusiastic tones and awkward pauses—OpenAI's latest update represents a significant step towards a more seamless and human-like interaction with technology.