Ruhi Sarikaya (Alexa AI director for applied science) outlined the advancements in machine learning technologies that have enabled Alexa to better understand people through contextual clues in a blog post released this week on the developer blog. According to Sarikaya, these enhancements have played a significant role in making Alexa more conversational and reduced user friction overall.
The self-learning technologies that Amazon has been using to train Alexa to automatically recover from its own mistakes have been in use since this winter. The system has been in beta testing up until now. It went live in the United States this week. Users’ “explicit or implicit contextual cues to detect failures of understanding or unsatisfactory interactions” are taken into consideration by the system, which does not require any human annotation, according to the researchers. It is possible to get contextual cues on anything from user preferences to previous behaviour to what Alexa skills are being used to what sort of device is being used and where it is being placed in the house.
Speaking of artificial intelligence assistances, Google Assistant may soon be able to provide its own note-taking and list-making capabilities, rather than relying on third-party programmes. The 9to5Google team has combed through the Google search app’s source code and discovered an undeclared “Lists and Notes” web application for Assistant that allows you to jot down important information that will sync across all of your devices. It is fairly rudimentary (you can’t do much more than include titles), but there is a significant level of polish that suggests it is more than just an experiment to be considered for publication.
It is currently unclear when or if Google will deploy this application into production mode. People may be perplexed by such a decision; after all, if you are currently using Google Tasks or Google Keep, why would you switch to something else? Despite the fact that it exists to assist users who do not desire to use stand-alone programmes, it can appear redundant if you are already deeply enmeshed in the Google bionetwork.
