The inventors of Apple’s virtual assistant software, Siri, have just demonstrated their secret, next-generation, artificial intelligence assistant. Created by Dag Kittlaus and Adam Cheyer, Viv is four years in the making. She offers an open platform through which queries can connect with third-party merchants. And yesterday, during Kittlaus’s demo at TechCrunch Disrupt NY, Viv performed flawlessly as her creator gave her tasks that would likely see Siri fumbling for answers.
Viv’s talent is in analyzing natural language queries, breaking them down to their components to determine intent, and feeding these requests off to bots that can easily process them. In this sense, Viv is a flexible “top bot” that commands a number of very inflexible specialized bots. This enables Viv to satisfy many unique requests by delegating tasks vertically. Whereas the highly specialized, personal assistant X.ai is capable of competently managing a calendar, Viv – like Siri – is designed to handle multiple tasks at various times. However, where Siri often falls back on a search engine to answer complicated questions, Viv seems to capably find an answer by relying on the support of the company’s partners. See Kittlaus’s demonstration and interview at Disrupt NY below.
When Kittlaus tells Viv “Get me a nice room in Palm Springs for Labor Day Weekend,” the system interprets the question’s intent and delivers the query to bots at Hotels.com. Similarly, when Kittlaus says, “I need a ride for six people from my office to Madison Square Garden,” Viv interprets that six people need a big car, so she connects Kittlaus to an Uber interface where he can book an XL or SUV. During the demo, Viv and its creator are able to perform four transactions in about two minutes, almost entirely through voice command.
Kittlaus and Cheyer are at the forefront of an natural language approach to AI assistants that includes competition from Microsoft, Facebook, and Amazon. But Siri and Viv’s creators aren’t newbies – they’ve been working with natural language comprehension for more than a decade. They do, however, still need to encourage merchants and developers to flock towards their software rather than that of the competition. To that end, Kittlaus demonstrated what he calls “dynamic program generation,” which allows Viv to display the code behind each command and query she processes. Kittlaus told Tech Crunch, “Instead of having to write every code instructed, you’re really just describing what you want it to do… The whole idea of Viv is that developers can go in and build any experience that they want.” With this code, developers can test and refine their user interface by speaking directly to Viv and editing sections of code directly. Kattlaus, Cheyer, and Viv’s investors hope that features like these will help the company build strong partnerships with developers, merchants, and subsequently users.
Image credit: Viv.ai