TECH

Google plans to make the Assistant's attention more selective in the future

Google has announced that future updates will make it more difficult to activate the Assistant by mistake. The company has also taken pains to reassure users that their audio data will not be collected by default in the same release. A specific opt-in determines whether the same will be analysed by humans or not.
Google has released a new statement in which it has outlined a commitment to avoid situations in which its Assistant is 'summoned' inadvertently, and hears conversations or other audio activity it shouldn't as a result. This is to be done by teaching the smart assistant the difference between overt activations (such as using the phrase "Hey Google") and things that only just may be the same.
The Mountain View company also took the opportunity to remind users that audio-data gathering practices will not be enabled by default, thus enhancing privacy in the presence of devices (in theory, at least). On the other hand, it appears that those who do have Voice and Audio Activity opt-ins switched on are liable to have their data listened to by human reviewers.
Google also now asserts that the audio data that is used in this way is anonymized, and that only 0.2% of all such collected data is used in this way. This new release may be part of a campaign to respond to a recent scandal in which the company was linked to the use of relevant data without user knowledge, and also possibly to maintain compliance with European data-security and privacy regulations.

by Deirdre O Donnell