Contextual Computing and the Next Era

TRANSCRIPT:
Hi, this is Tim Tuttle. I'm the CEO of Expect Labs. This is another installment in my video blog series about the ways in which intelligent assistants are going to get smarter over the coming years. Now my most recent video blog was all about how intelligent assistants are going to get smarter, much smarter, in the coming years because of the expansion of the underlying knowledge graph that these intelligent systems are built on. And I posed in my previous video that the knowledge graphs that we're going to have at our disposal, 5 to 10 years from now, will start becoming so large that they will begin to approach a large percentage of all human knowledge. What I said in the previous videos is that those knowledge graphs would probably need to have potentially trillions of concepts, which by the scale of Internet data bases today is certainly within reach. But that's only half of the problem, right? Or half of the solution.
Say you have access to a knowledge graph that contains nearly every concept known to the human kind. The big problem then becomes, how can you search and find just the pieces of that knowledge graph that are relevant to you? Let me give you an example. Let's say, you want your intelligent assistant to understand when you say something like, "I met my friend Joe at Luna Park yesterday." Well, it does no good for an intelligent system to know that there are 45,000 people in the world named Joe, and it does it no good to know that there are 1,500 restaurants or businesses that might be called Luna Park. What really matters in that sentence is that it know that you have a friend Joe and knows which person you're talking about. It also knows that there's a restaurant named Luna Park that's right near your house that you go to frequently. In order to understand that, what is required is an understanding of your own personal context.
What is your own personal context? Well, your personal context in essence is all of the concepts, all of the subjects, all of the entities, that you have knowledge of right now at this point in time, that you may be potentially referring to in a question or a conversation that an intelligent assistant is trying to understand, Well how big is that personal context? How many concepts does an intelligent system need to understand in order to be able understand your personal context? Well, you can get an estimate of how big your personal context is by essentially looking at all the words that a person might have in their lexicon. And this is a pretty well understood concept you can look at instead of the language that people use. For the most prolific writers, the people with the largest vocabularies, they typically have tens of thousands of words at their disposal in their own vocabulary at any one time. So they tend to have as many as hundreds of thousands of words that are in their head and this is the extent of human's ability to have that set of data in your head. So if you use that as an analog for trying to capture what are all the known concepts that a person might have in their head at any point in time, it's probably in the tens of thousands or maybe, low, hundreds of thousands. And what are those concepts? Well, so what your personal context is is the name of any friend or person that you know or have ever met. The name of all the locations and places of interest that you have been in your recent history or that are currently nearby. It can include any of the concepts have read in a book or in a news article, or any things or people or concepts that you might have seen in a video or movie that you might have. That is your personal context. it's actually the collection of concepts and entities that are in your brain at any one period of time.