[kazinator]

How about having the AI keep talking until you say a specific word like "stop".

Background noise will rarely produce a false positive for that word.

[KeplerBoy]

Because that's a completely different problem. By doing that you'd reduce the problem to transcription and have a worse UX.

[YetAnotherNick]

They are solving the opposite problem, ie detecting when to start the AI. Stopping the AI is significantly easier problem.

[kazinator]

In spite of your assiduously numerous re-readings of the fine article, it appears that the following passage has eluded your attention:

As a result, the VAD mistakenly interprets noise or background voices as active user speech, triggering unintended interruptions. These false triggers negatively impact turn-taking, a core component of natural, human-like conversational interactions.