How often have you felt frustrated by Siri, Cortana, Alexa, Google Assistant, Sher.pa or other Digital Assistants?
We were promised a talking, feeling, engaging Digital Assistant in films like Her or Apple’s 30-year-old video Knowledge Navigator envisioned by then-CEO and my friend John Sculley. But the actual user experience falls tremendously short of those promises. And, sadly, it will take years or even decades before we see a Digital Assistant that reflects those fictional portrayals.
Although they have a wide range of capabilities, today, the harsh reality is that Digital Assistants are used typically to tell us about the weather, look up simple information, set an alarm, or tell us a joke. They have other tricks, of course, if we learn what those tricks are. We have to learn what our Digital Assistants are capable of doing. It shouldn’t be like this. We shouldn’t have to learn about our Digital Assistants; they should learn about us.
Digital Assistants can look up answers to millions of questions and execute specific tasks. But the human mind is capable of imagining literally billions of possible combinations of questions. It is easy for the human mind to skip from one set of capabilities, say book a table at a restaurant, and assume the same skills can be used to perform another task, for example, reserve a ticket for a sporting event. But just because we can imagine that it ought to be possible doesn’t mean the Digital Assistant can actually perform the task.
That’s where the disappointment comes in. We have reasonable expectations, based on our very human capabilities, that are impossible to be fulfilled by today’s approach to Digital Assistants. Some years ago, I discussed this point with Tom Gruber, the co-founder of Siri.
So why don’t we have a human-like Digital Assistant yet? Because we start with what the technology can do, rather than focusing on what the user experience should be. We build capabilities that the technology can do, rather than capabilities the product should do. We must rethink Digital Assistants from the user’s point of view, leverage technology, and build a product that delivers the features a user needs without over-setting expectations.
The mistake is understandable. Apple, Amazon, Google — even my company, Sher.pa — have made the same mistake: We have built Digital Assistants only as automated, voice-enabled search and task interfaces. They are systems that respond to specific questions or voice commands, such as “find information” or “turn on the lights”.
But our expectation is that these systems should mimic a Human Assistant.
Three years ago, my company paused for reflection. The product we called a Digital Assistant wasn’t much of an assistant, at all.
I started to think about Martin, the person who works as my Personal Assistant. The truth is, I don’t ask him to tell me jokes, turn on or turn off lights, set alarms, or answer questions about the weather. I need much more from him. I need him to learn about me, to know what I like and don’t like, for him to understand the context of my day and to manage my time to help me be well informed with relevant information even before I ask. And I want him to do this without my having to tell him much at all. Martin is a great Personal Assistant because he is proactive and anticipates my needs.
None of today’s Digital Assistants have anything close to Martin’s capabilities. That needs to change. A Digital Personal Assistant that is more human-like would surely be a big step toward the vision that Apple expressed so many years ago. That is the vision that consumers should demand.
So, we started from scratch to build a next-generation of Digital Personal Assistant that actually worked more like a Human Personal Assistant. We have started with a clear vision of the relationship between the user and the assistant, and our AI team is building our next generation product, just as we hope other companies are building more user-centric systems.
In 2018, we’ll see the first versions of this next generation of Digital Personal Assistants, from my company and from others. Our challenge is to make them at least as helpful as Martin is to me.