At the Alexa Conference last week, there was a shared anecdote about the Echo becoming more prone to false trigger (where it detects someone saying “Alexa” even when they haven’t). The audience was polled and at least half noticed that the Alexa wake word had become more sensitive. I thought about this again when my home Echo triggered against something my three year old had said.
While there wasn’t any evidence pre/post any time period to verify false trigger rate provided at the conference, it is possible that Amazon made the local trigger more sensitive after it implemented Cloud Verification of its trigger word. This technology sends both the command but also the wake word that’s buffered by the mic to Amazon’s servers to verify that indeed a wake word was spoken. If no wake word is dedicated at the beginning of the stream, Amazon shuts down the stream.
While privacy advocates might freak out that Amazon is always buffering an audio recording (even if it’s not sending) and the wake word might be more sensitive and accidentally stream audio, the counter argument is that Amazon is more likely to shut down a false trigger stream preventing unintended commands from being recorded. Most people won’t notice an increase in false trigger (it just lights up the device and then the device’s lights shut down).
Perhaps there might be a backlash at some point against the increase wake word firing and cause an increased focus on even better local wake word implementation. However, it might take a huge security breach for that backlash to get unleashed.