NEWS ANALYSIS: The day is coming when some people have so many virtual assistants they will need a virtual manager to direct their "virtual staff." But how many virtual assistants are too many?
The availability and use of digital virtual assistants—software-based artificial intelligence services that do things for you—is about to explode.
Today, we think of the all-purpose assistants like Google Now, Apple's Siri, Microsoft's Cortana, Facebook's M and Amazon's Alexa. To some degree, we feel compelled to choose which we'll use.
Sometimes that choice is easy. If we're Apple fans and iPhone users, then we'll probably choose Siri. If we're Google fans and Android users, we'll probably choose Google Now. No matter what, most current users prefer one over the others and tend to see it as an either/or choice.
There's a surprising business benefit for the big companies to create highly complex virtual assistants. For example, Siri was a stand-alone iOS app before it was a core feature of iOS (that was before Apple bought the startup that created Siri).
When Siri was a mere app, it functioned in only three categories—restaurant reservations, travel and movie tickets. Now, Apple has turned it into an all-purpose app that gives you the weather, does currency conversion and wittily answers questions about the meaning of life.
Why did Apple want Siri to do everything, rather than just three things? And, for that matter, why do all the major tech companies—Google, Apple, Microsoft, Facebook and Amazon—all insist on creating virtual assistants that are generalists, rather than specialists?
In fact, these assistants, with the exception of Facebook's M, also involve the incredibly complex feature of automated voice recognition. Google Now, Siri, Cortana and Alexa all have to understand human speech, find the best answer and deliver it with a humanlike reply. Facebook's M needs only to understand text, which still requires some complex processing. These services involve voice recognition, artificial intelligence, vast databases of knowledge, personalization and natural language replies.
These virtual assistants are highly complex, difficult and expensive for these companies on multiple levels. The benefit of this breadth and complexity is that no small startup could ever do it.
With an explosion or revolution of virtual assistant services and products just around the corner, the big, deep-pocket companies are relying on the cost and complexity of their products to thwart potential competition from innovative, new assistants created by startups.
What startups are doing in ever increasing numbers is launching specialized virtual assistants that function in a more limited way or deal with a narrow category of information.
At the current rate of development, we can expect to see several specialized virtual assistant products or services each for banking, insurance, health care, law enforcement, law and other professions.
Most of these will be built by startups, and many of these startups will build their products with open APIs offered by the major artificial intelligence companies. Similarly, we can expect to see these companies subscribing to huge datasets or using free and open APIs. They'll leverage third-party voice recognition engines and other components.
Virtual assistants are not just for professionals. Consumers will get virtual assistant platforms exclusively focused on cooking, shopping, fitness, self help and psychology, home automation, cars, travel and more.
A company called ToyTalk
already created separate virtual assistant platforms for Hello Barbie and Thomas & Friends Talk to You platforms—two virtual reality systems for children only, and each different from the other.
There are even companies working on the idea of enabling users to build their own artificial intelligence agents. A company called Arya
offers roll-your-own virtual assistants, but requires you to supply your own data.
We can also expect to see virtual assistant features built into apps as an interface. The main point will be the app, but the virtual assistant features will function as the only way to interface with that app.