速递｜为硬件铺路：OpenAI攻坚下一代音频AI模型，打造“主动式”系列AI设备

Core Insights - OpenAI is enhancing its audio AI models in preparation for the upcoming AI-driven personal devices, which will primarily utilize audio interaction [1][2] - The first of these devices is expected to be released in the first quarter of 2026, with ongoing efforts to improve the accuracy and responsiveness of the audio models [3][4] Group 1: Audio Model Development - OpenAI has integrated multiple engineering, product, and research teams over the past two months to optimize the audio model for future devices [2] - The new audio model architecture is reported to generate responses that sound more natural and emotional, providing more accurate and in-depth answers [2] - The new model can also handle simultaneous speech with human users, a feature that current models lack [2] Group 2: Device Features and User Interaction - The upcoming device aims to serve as a companion, proactively offering suggestions to help users achieve their goals, rather than merely acting as a conduit for applications [7] - OpenAI is exploring a series of devices, including glasses and display-less smart speakers, rather than a single device [5] - The design philosophy emphasizes voice interaction over screen-based communication, as many researchers believe that speaking is a more natural way for humans to interact with AI [3][4] Group 3: Challenges and Leadership - A significant challenge for OpenAI is that many ChatGPT users do not currently interact with the chatbot vocally, which may be due to the low quality of the audio model or lack of awareness of the feature [4] - Key figures driving the audio AI initiative include Kundan Kumar, who leads the project, and Ben Newhouse, who has adapted OpenAI's infrastructure for audio AI [4]