To Be Truly Useful AI Assistants Need To Learn To Anticipate

As AI-powered automated personal assistants have increasingly found their way into our lives, their tremendous power at certain tasks has often been undermined by their inability to fuse the data available to them into a comprehensive view of our lives and, perhaps most importantly, their inability to actively anticipate our needs rather than merely passively respond to queries posted to them. How might adding proactive anticipatory reasoning and the ability to look across data allow our future AI assistants to be far more useful in our day-to-day lives? As I was taking a cab to the airport this past Friday, I noticed everywhere around me preparations for the Marine Corp Marathon, which I had completely forgotten was this weekend and which meant that when I returned early Sunday morning I was going to have difficulty getting home given all of the road closures in my neighborhood. Yet, despite having access to my calendar, which clearly noted my return flight arrival on Sunday, and being able to tell me that there was a giant marathon running directly through my neighborhood on Sunday with road closures all around my home, my AI assistant was unable to connect the two and anticipate that I might have trouble getting home via my usual route on Sunday. Heading to a meeting a week ago, my assistant could tell me there was a huge traffic delay along the way when I explicitly asked for a traffic update, but was unable on its own to connect that to my next calendar appointment and proactively suggest 30 minutes earlier that I leave half an hour early to avoid being late.