Social RoboticsOpenAI's GPT-4o brings near real-time voice conversation
A single model handling voice, vision and text let users talk to an AI that responds almost as fast as a person.
OpenAI launched GPT-4o, with the o standing for omni, combining speech, text and vision in one model instead of chaining separate systems. Live demos showed it translating between languages, tutoring and reading stories in different styles, while letting users interrupt it mid-sentence. Reported voice response times were close to those of human conversation. For social robots, which have long struggled with stilted turn-taking and lag, low-latency expressive voice AI was a major shift in what conversational machines could feel like.
Source: MIT Technology Review
This archive entry summarises reporting published elsewhere; follow the source link for the full story. The summary was drafted with AI from that reporting.