Let’s say, for example, you want to keep track of that pollen level. You would navigate to the website where that information exists, click on the specific data you want to track, then save the page as a pin within the assistant extension.
The pollen level will then display on your personal assistant dashboard, which will update as new information is available. It might be one of a dozen types of information your personal assistant tracks and displays, from package tracking to unread emails to campsite availability. And you can ask to be notified if that pollen level hits a certain value or have your assistant automatically book you a campsite and let you know.
Those notifications could come on the browser itself or via a not-yet-built mobile app, made possible because the browser is running in the cloud and uses WebDriver to simulate basic browser navigation such as clicking and scrolling. For tasks like checking email, the user’s cookies would be securely passed to the cloud and used to refresh the data.
But, of course, your new assistant is just getting to know you and your needs, so it may not get everything right on the first try.
That’s why another critical piece of the project is a user’s ability to search for information using their own natural language, and to continuously provide feedback to train the personal assistant to become more useful over time.
So, if you type “What is the ragweed pollen level today?” into the search bar on your personal assistant dashboard – or, eventually, ask it out loud through your app – your personal assistant would display what it thinks is the information you’re after. If it gets it wrong, you tell it.
Over time, the personal assistant not only gets more accurate, but learns your specific style of searching.
“The trainability of this model is really compelling,” said Jeff Ramos, who heads the Microsoft Garage. “It could really set Edge apart in the browser space and even automate tasks across platforms.”