To be truly useful AIs need to learn who we are

Siri, Cortana and other personal assistant apps are just the beginning of how AI-based systems will help us in our daily lives, but in order to be truly useful an AI needs to learn more about us.

Asking your smartphone to tell you where the nearest pizza place is located may seem like a piece of cake (or a slice of pie, in this case) but it involves a lot of very sophisticated technology. First, you need a smartphone and there is a ton of tech in smartphones. Next you need voice recognition – another trick that we take for granted these days although it is an incredibly difficult thing to do. Next you need GPS location capabilities, and at the very least you need satellites to do that. Finally, you need nearly instant access to massive databases that contain maps and information about pizza place locations.

Still, that’s a handy little trick that wouldn’t have been possible a few years ago. But AIs need to get a bit smarter if they want to become truly useful for more than just finding fast food joints.

Some of the creators of Siri are working on a next generation AI called Viv that should be able to do more. Once completed, it should be able to understand more complicated instructions so, for example, you might be able to say, “My daughter wants to order pizza for her sleepover party, can you text her to see what toppings everyone wants then place the order so I can pick it up on the way home?”

Then perhaps someday it could check which of her friends are expected to show up and make sure no one is allergic to pepperoni.

I also believe that AIs will have to learn who we are and customize themselves over time. I don’t necessarily think they need to control our lives (unless we want them to). I don’t want my personalized AI to balk if I order a pizza (‘You’ve gained five pounds in the past month, Guy, perhaps you should order a salad instead.’)

But I wouldn’t mind so much if I specifically asked for advice, “If I get a small pepperoni pizza is that going to completely blow my diet?” or “I really want a slice of pizza for lunch, could you remind me to skip desert tonight.”

Or even offer advice that takes into account other parts of my life. “I really want this flashy new flatscreen TV, is there any way I can work it into my budget?” And if it came back and said something like “This particular model is out of your price range and it didn’t get very good reviews anyway. There is a different model at another store less than a mile from here that received much better reviews and it’s half the price,” that would be truly useful.

Letting our AIs have access to more of our personal lives could make them infinitely more useful (and more scary) but they could also help protect us from ourselves and others. I can easily imagine my AI warning me that someone was trying to access my email account or is using my credit card for unauthorized purchases. Or if I lost my smartphone and someone else tried to use it my AI would realize it’s not me and would lock down all my data and disable the device.