The presenter describes reported full-duplex voice interaction and the ability to delegate deeper tasks while a conversation continuesAI agent delegation assigns a bounded task, context, authority, and expected result from one participant to an agent or another agent.. He compares published benchmark figuresA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions. with earlier voice modes, presenting them as reported measurements rather than independently verified results.
Demonstrations illustrate discussing imagesA multimodal model can process or generate more than one kind of data, such as text, images and audio., retrieving informationRetrieval is the process of selecting relevant stored information and returning it to an AI system for the current task. during a conversation and translating between languages. The presenter views these as examples of a more interactive voice interfaceA voice interface lets people interact with software through spoken input and audible responses..
A final demonstration concerns conversational timing and short reminders. These are selected examples of intended behavior, not evidence that every interaction or background task will work reliably.
Watch on YouTube
