Apple researchers are investigating the use of AI to identify when a user is speaking to a device without requiring a trigger phrase like 'Siri'.
A study involved training a large language model using speech and acoustic data to detect patterns indicating the need for assistance from the device.
The model showed promising results, outperforming audio-only or text-only models as its size increased.
Eliminating the 'Hey Siri' prompt could raise concerns about privacy and constant listening by devices.
Apple's handling of audio data has faced scrutiny in the past, leading to policy changes regarding user data and Siri recordings.
[link] [comments]