OpenAI 重組團隊以開發基於語音的人工智慧硬體產品

OpenAI reorganizes some teams to build audio-based AI hardware products

Voice has lagged in adoption behind screens. OpenAI wants to change that.

OpenAI, the company that developed the models and products associated with ChatGPT, plans to announce a new audio language model in the first quarter of 2026, and that model will be an intentional step along the way to an audio-based physical hardware device, according to a report in The Information.

Citing a variety of sources familiar with the plans, including both current and former employees, The Information claims that OpenAI has taken efforts to combine multiple teams across engineering, product, and research under one initiative focused on improving audio models, which researchers in the company believe lag behind the models used for written text in terms of both accuracy and speed.

They have also seen that relatively few ChatGPT users opt to use the voice interface, with most people preferring the text one. The hope may be that substantially improving the audio models could shift user behavior toward voice interfaces, allowing the models and products to be deployed in a wider range of devices, such as in cars.

OpenAI plans to release a family of physical devices in the coming years, starting with an audio-focused one. People inside the company have discussed a variety of forms for future devices, including smart speakers and glasses, but the emphasis across the line is on audio interfaces rather than screen-based ones.

The company is not alone. Its competitors, including Google, Meta, Amazon, and others, have increasingly turned their R&D efforts toward products and technologies that are focused on voice-and-audio interfaces, like Meta’s push into smart glasses.

This is, of course, not the first time such a push has happened; there was the boom of Alexa, Google Assistant, and (to a lesser extent) Siri-driven voice assistant devices just a few years ago. Those assistants have been relatively popular with some subsets of users—typically casual tech consumers rather than hardcore technologists.

However, the devices had significant limitations. New approaches based on large language models may open up new possibilities (and risks).

Some developers of AI products—including former Apple design lead Jony Ive—believe that voice-controlled products are likely to be less addictive than screen-based ones and have stated that as a reason for pursuing audio interfaces, though they have not generally cited reliable evidence of this.

OpenAI’s first audio-focused physical device is currently expected to ship about a year from now, but we don’t yet know much about what it will look like.

Ars Technica has been separating the signal from
the noise for over 25 years. With our unique combination of
technical savvy and wide-ranging interest in the technological arts
and sciences, Ars is the trusted source in a sea of information. After
all, you don’t need to know everything, only what’s important.