Called “personalized voice recognition”, it is the new frontier of Google’s artificial intelligence. This is a feature that will soon be available in Google’s app for Android smartphones, and is linked to Google Assistant.
This feature was discovered by 9To5Google in the app code and is not yet active, so any changes have yet to take effect. But Google is already working on it, so much so that the new feature is already in the app’s code. But that’s not all: in the same code, there is also the option to deactivate the new function, a sign that Google will let the user choose to use it or not. But, at the same time, it warns the user that disabling personalized voice recognition will make Google Assistant less accurate.
Personalized voice recognition
The new feature, as we have already mentioned, is not yet available, but we already know how it will be described by Google: “Record audio recordings on this device to help Google Assistant improve the recognition of what you say”. The audio remains on this device and can be removed at any time by disabling “personalized voice recognition”.
According to 9To5Google, the technique that will be used to improve voice recognition will be “federated machine learning” which Google says maintains a high level of privacy but, at the same time, helps reduce erroneous voice assistant activations.
Too often, indeed, it happens that Google Assistant (but it also happens to Amazon Alexa, to be honest) activates incorrectly because it thinks it heard the words “OK Google” or “Hey Google”.