OpenAI has introduced GPT-Live, a new voice model for ChatGPT that allows real-time, simultaneous listening and speaking, making AI conversations feel more natural. During a Wednesday livestream, the company demonstrated the model's ability to acknowledge users mid-sentence and handle multiple requests at once.
TL;DR
- OpenAI launches GPT-Live, a voice model enabling real-time AI conversations.
- The model allows ChatGPT to listen and speak simultaneously, acknowledging users mid-sentence.
- This update makes AI interactions more natural and competitive in the evolving AI voice assistant market.
What happened
During a Wednesday livestream, OpenAI unveiled GPT-Live, a new voice model designed to power ChatGPT Voice. This update allows ChatGPT to listen and speak simultaneously, rather than waiting for the user to finish speaking before responding. The model also acknowledges users with intermittent verbal cues like 'mmhm,' 'yeah,' and 'got it' while they are still talking.
In a demonstration, a user asked ChatGPT to fact-check a meeting date while also checking the weather and traffic along their route. ChatGPT responded with short acknowledgments and continued working through the request as the user added more information. OpenAI also showcased ChatGPT's ability to translate language in real time, keeping pace with the conversation as it unfolds.
Why it matters
GPT-Live makes AI interactions feel more natural and less like a baton-passing chatbot. This update addresses common criticisms of AI language models, such as their awkward pauses and overly chipper tone. By enabling real-time conversations, OpenAI aims to create a more seamless and engaging user experience.
This launch positions OpenAI competitively in the evolving AI voice assistant market. Other companies, like Thinking Machines, are also working on similar technology to make AI interactions more fluid and natural. As AI language models become more integrated into daily life, the ability to handle real-time, continuous input and output will be crucial for user satisfaction and adoption.
Key facts
- GPT-Live is a new voice model powering ChatGPT Voice, enabling real-time, simultaneous listening and speaking.
- The model acknowledges users mid-sentence with verbal cues like 'mmhm,' 'yeah,' and 'got it'.
- In a demo, ChatGPT handled multiple requests at once, including fact-checking and providing weather and traffic information.
- OpenAI showcased ChatGPT's ability to translate language in real time, keeping pace with the conversation.
- Greg Brockman, OpenAI's president, described the update as a 'much more natural way of interacting with your computer'.
- Thinking Machines, led by former OpenAI chief technology officer Mira Murati, has also teased similar technology for handling continuous input and output across audio, video, and text.
Context
The launch of GPT-Live comes at a time when AI language models have faced criticism for their awkward interactions and unnatural pauses. As these models become more prevalent, companies are focusing on making AI interactions feel more human-like and seamless. OpenAI's update is a step towards addressing these criticisms and enhancing the user experience.
The AI voice assistant market is evolving rapidly, with companies like Thinking Machines also working on similar technologies. As AI becomes more integrated into daily life, the ability to handle real-time, continuous input and output will be a key differentiator for AI assistants. OpenAI's GPT-Live positions the company competitively in this growing market.
