Is Gmail’s Priority Inbox A New Kind of AI Personal Assistant?
Google announced Monday that it would be rolling out a new email sorting feature for its 170+ million Gmail users. The "Priority Inbox" allows you to train Gmail how to rate incoming email according to its importance. Those messages that you actually need to pay attention to are bubbled up to the top so that you can find them easily. According to Google, Priority Inbox judges the importance of emails by monitoring which types of emails you regularly read and which you reply to. You can also individually mark the importance of an email through plus and minus buttons or set up filters to automatically mark messages for you. Along with Gmail's evaluation of spam (which is quite good), Priority Inbox has the potential to manage even the worst email overload. Watch Google's cutesy ad for the beta product below. I think programs like Priority Inbox are the key ingredients in the next generation of artificial intelligence enabled personal assistants.
When AI personal assistants were discussed in commercials in the 90s, companies predicted that they would be able to do almost anything: pay your bills, drive your car, defeat your enemies with real kung-fu grip, etc. While our current generation of AI assistants has yet to reach these lofty goals, they have become very adept at one function: relating your commands to other programs. Google's new Voice Actions for Android Phones, and Apple's newly aquired Siri for iPhones, let you speak commands and rely on the assistant to manage other applications to get the job done. In short, our AI assistants have become fledgling middle managers, handling our outbound requests. Priority Inbox, however, displays an example of another important quality of a personal assistant - the ability to manage inbound information.
Priority Inbox is probably very low on the AI food chain. It's little more than a smart filter than can learn. Yet that sort of ground-level information management is an essential part of what we will want AI assistants to accomplish. We're are going to be flooded with incoming information as billions of people around the world jump online in the next few decades and begin to rely more and more on electronic communication. Which phone calls will you answer? Which Facebook friends do you really want to stay in contact with? Which news feeds will you actually need to read? AI personal assistants will be able to manage this information with simple filtering, allowing these data streams to enhance your experience rather than overwhelm it.
Google has already discussed other data management AI projects (Big Query and the Prediction API) that will allow companies to apply even better analysis to the information they gather. These programs are part of the same basic ecosystem - they manage incoming data so that you know what's important, and what actions are most profitable to take next.
I think what we're seeing is the slow formation of the two paradigms that will constitute the next generation of wide spread AI. We'll want learning programs to manage incoming data, and we'll want them to translate our commands into outgoing actions. These narrow applications of artificial intelligence are pretty much already here, but they are bound to get better. I don't look into my spam folder anymore, I trust Gmail's filters enough that I don't feel like I have to double check them. With sufficient training, I'm sure I'll end up trusting Priority Inbox in the same way. Eventually we'll have analogous programs for all forms of digital communication: calls, txts, news feeds, etc. I'm slowly getting Siri to where it understands my commands well enough that I don't double check it for some tasks. Eventually personal assistants will be trusted to manage all our mundane digital requests. If you told me fifteen years ago that I would be happy to have a virtual assistant that told me what information was important, and managed all my commands to other programs, I would have assumed that the world have been taken over by machines and that I was being doped with soma. Now I think such programs would not only be a welcome addition to our digital world, I believe they are halfway here.