By Kirk Douglas
Dag Kittlaus took the stage on Monday May 9th at the New York City TechCrunch Disrupt conference to unveil what his company believes to be the beginning of a revolution in Intelligent assistance software. Viv, (pronounced like the first syllable of the name Vivian) actually stands for Life. And as the CEO denoted, the name was decided upon because it embodies his company’s mission to breath life into the current state of voice recognition and artificial intelligence.
Kittlaus is no stranger to the technologies and science required to achieve such an extraordinary feat prior to starting his efforts with Viv, he was a co-founder of the Apple-acquired Siri intelligent assistant. After selling to Apple in April of 2010, Kittlaus moved onto his new project in search of ways to expand upon the usefulness and fluidity of interactions between humans and artificial intelligence.
In his keynote presentation of the Viv software on Monday, he revealed in detail and with a great number of example demos, how his company is planning to change the way we interact with our digital assistants.
He gave what I would consider a very palatable explanation of both the company’s intent with the software and how it works. In fact, “intent” – the word itself – was a large part of the theme throughout his discussion as he described the scientific breakthrough he believes his company has made in understanding human language.
Kittlaus described a software system Viv is built around which is dynamic. It uses Nuance’s voice recognition to determine what is being said and then derives from that string of words what the user is trying to achieve. This, as he calls it, is the “intent.” Once the intent of the input is determined by the Viv software, it can analyze a network of proposed solutions and “write its own program” in less than a 10 millisecond time frame to execute appropriate results for the end user. In other words, his product is aiming to understand what the user is saying. His product is aiming to understand the actual intent of a user when asking for something rather than just playing a limited game of match-up.
Dag pointed out that this is critical to making a better digital assistant because current technology in use is based off a system of program management. Current A.I. assistants (like Siri and Google Now) reach results by way of hard-coded proposed inputs that are matched to their appropriate outcomes, also limited in scope and also hard-coded. This makes them inherently limited in proposed solutions they can provide to the end user. He went on to explain that making Viv dynamic and setting up a constantly growing network of API’s and partnerships means that his software will be able to achieve a scale that we have never seen before. He explained that growing partnerships and inviting developers to provide unique solutions will make the Viv system smarter and allow for all kinds of interactions moving forward that we haven’t even thought of yet.
The end goal, as he stated in a follow-up Tech Crunch interview, is to achieve platform ubiquity. Kittlaus clearly expressed that Viv is a solution that he would like to see through and though his company has already been approached about potential acquisitions, he held firm that he will not consider any option that doesn’t suit that planned goal for ubiquity.
His presentation seemed about as well-received as anyone could ask for given the positive write-ups in the tech-sphere’s postings shortly following his presentation. And rightfully so in my opinion.
Kittlaus started with what most of us would expect, a simple question asking what the weather at home was like. Results populated quickly from an integration with Weather Underground that Viv was able to pair up with, presumably through a combination of that partnership and known geolocation info regarding where home is. This however, was just the start. To show off the breakthrough in intelligence prowess, he threw increasingly complex demands at Viv.
In one example, he ordered flowers for mom on her birthday and in just a couple of taps they were ordered and set for delivery. In another, he requested a ride for 6 people from his office to Madison Square Garden. Watching the results come back for an Uber XL with awareness of both locations was an achievement in itself by today’s standards.
In perhaps the most compelling (though not the most realistic) of examples, Kittlaus asked “will it be warmer than 70 degrees near the Golden Gate Bridge after 5 PM the day after tomorrow.” To which the software responded, “No, it won’t be that warm Wednesday after 5 PM.”
Impressive technology indeed. His product demonstrations wowed in their increased complexity and contextual awareness, making it clear that Viv is treating verbal data in a much different way than the widely deployed ‘identify and match’ A.I. that has become commonplace as of recent.
Viv is set for a wide release cross-platform on mobile devices by year end 2016. Kittlaus pointed out that it’s developer platform for Viv will release first with select partners and later to the mobile developer community at large.
I watched the extended announcement on TechCrunch including a follow-up Q&A and suggest you do the same if you find the future of mobile A.I. to be exciting. (Video below)
It makes sense that if A.I. in personal assistants is to reach critical mass, the architecture required to get it there will need to be built around understanding human intent. The approach Kittlaus is taking seems to have serious potential given the right partnerships and the scale that might come with a ubiquitous approach. The success of Viv without being tightly coupled to an OS could certainly present problems for the company moving forward, but those troubles remain to be seen. So far, I haven’t seen anything quite as compelling. Viv is by far the most intelligent assistant that I’ve ever seen demoed and its existence serves to re-ignite the dream of a bright future for A.I.