Thank you. Good afternoon. so, what I'll be talking about is how we can actually get to what we heard about this morning level 4 and level 5 of conversational Ai and how we can actually do that what needs to be done. But let me start off by giving you a little bit of context my background. I've built several technology companies over the years and I've al so, developed several technology platforms including Programming languages and database systems also, Erp system as you might yell P company was very successful when from the garage to 400 people and did an IPO. So, that was very nice. It's really that it enabled me to say what is the big problem that want to tackle that bothers me and what struck me is that software is really stupid and I'll say that, you know being very proud of the software that I've developed software can't think and reason I can't learn it doesn't have common sense. So, how can we solve that problem? And so, I spent five years doing my own research and really studying up on what intelligence entails. What is cognition? What do IQ tests measure how do children learn how does show that our intelligence different from animal intelligence and of course what had been done in AI to try and build thinking machine? Ins so, over that five-year period I came up with a designed with various design ideas to do that. And I then formed a company an R&D company in 2001 210 turn my ideas into actual code into platform. so, for several years for another five years. We were just an R&D mode basically figuring out how to build thinking machines in 2001. I actually al so, coined the term artificial general intelligence to There was two other people and we wrote a book on the topic and it's kind of nice to see that this term is now quite commonly used to refer to truly thinking machines, So, by 2008 we had enough of the technology together that actually launched the company in the ivr space automating calls in the call center that you know, as we've heard many times. We all know people hate it when they call into a company and they have to talk to computer. They just press the arrow to get to an operator. Well, we provide a much better experience with Smart Action. You can speak in natural language right through the Patients are can start off by saying how can I help you? Right through the conversation understands natural language and we've been doing that for 10 years. Now, if you break down the side of the road new AAA member chances are our first generation, I go will actually help you get going again. However, I've found that I exited the company five years ago to really focus on taking the technology to the next level. To be able to handle wider range of conversations and more complex conversations, so, there is huge demand for truly intelligent personalized conversation both as a user interface to devices and al so, as a personal assistant so, you know with as we've been talking to customers, we see there's tremendous. Of interest in all sorts of different areas, you know, some of the obvious ones are the help desk in education. If you have a robot and a hospital a hotel you obviously want to be able to talk to it and you're expected to understand you Automotive, you know, you want to be able to talk to your car and expected to understand you VR. Are very obvious applications where you need voice interaction, but you need something that understands you in the middle that refers to sort of a co-pilot application to be able to talk to software that you can just talk to your Oracle sap or whatever software were your sales for software and say, you know, give me the sales for the last three months by product in Europe and it should be able to understand you and then say always shows us to me as a as a pie chart. And should remember that it would al so, have a memory that you can say run the cash flow report you did for me last week and it might then come back and say I got my come back and say do you mean the one for Nevada or the one for California? I think there's tremendous demand for this kind of thing. However, we know there yet. Why don't we have that now but handle simple applications really well as we've heard quite a few applications that are running well, but there Al so, really serious limitations and to concretize it. We I have a little snippet here of the market leader generally recognized as a market leader Alexa trying to do some relatively straightforward things. All right, Alexa, add chips and guac napkins and three bottles of wine to John's birthday list. I added chips guac napkins three bottles of wine to your shopping list. Actually, Alexa remove the napkins from the list to clear your list or remove an item visit the Alexa app. Okay, then. Alexa what do I have on John's birthday list you have one item on your shopping list chips guac napkins three bottles of wine. You call all that one item. Alexa remember that John's favorite cake is red velvet. I'm not quite sure how to help you with that. Okay, I'll talk a little bit more about what the limitations are there. Now, I'll show you our technology hooked up to the Alexa microphone speaker creating John's birthday list. Is that okay? I added chips and guacamole napkins and three bottles of wine to John's birthday list. What was the food on that list again chips and guacamole and three bottles of wine? Actually, let's make that five bottles. Sure. The list is now five bottles of wine chips and guacamole napkins. Great. Let's send it over to June and ask her if there's anything I'm missing please sending John's birthday list to June via email. Wait, always contact June by Skype. Okay sending the list to June via Skype. Remember that John's favorite cake is red velvet. Got it. fantastic What a mock-up, but I also, didn't show you the blooper reel. so, I go if I doesn't understand something it will ask. What is guac the same with guacamole. It will now learn that and remember that and you know, it can reason about things. What are the food items on the list and use context to say make that five bottles and then learn simple things like always contact you in by Skype and John's favorite cake is red velvet? This is very general capability to be able to learn and reason and disambiguate on the Fly. Now why don't we have this technology available? Generally, and I generally say its kind of an accident of history that the current machine learning technology has been so, successful that is sort of sucked all the oxygen out of the air and that's what everybody's working on. Let's just throw more data at it more processing power. It's not going to solve the problem because the current chatbots offered by all the major suppliers plus the hundreds of almost all of the other small companies offering them really suffer from severe limitations cognitive limitations. They don't remember what you said five minutes ago. Never mind what you said last. Week they cannot learn interactively unless they were specifically programmed to listen for something. They don't have deep understanding because they just do statistical pattern matching, so, if you say I hate you, but don't ever give me Uber again chances are it will still trigger Uber the Uber app. so, there's no reasoning so, you can't really have a conversation and they're not personalized now. The reason for that is we can find that What topicals? the third wave of AI now the current technology that everybody almost everybody is using essentially is a categorize a big fat categorize ER you'll collect a lot of data you tag it you train the categorize ER and then you write a little flowchart E-Type programs essentially to have the conversation flows and you know, the limitations are fairly obvious was that approach now the To overcome that is to really move to a different approach of technology. So, the topic talks about these three waves of AI first wave is good old-fashioned AI basically logic programming expert system. so, on we know we can't overcome, you know, we can't handle conversation through flowcharts alone. so, then the second wave, you know promise to kind of solve that and It has been tremendously successful in a number of areas and speech recognition image recognition and so, on and you know taking a lot of data and basically building statistical models from that. However, it's Al so, become clear that the limitations of that are quite fundamental and severe to the extent that the Godfather of people earning year or so, ago said My Views to throw it all away and start over even the head of a deep mind. Just a few months ago said deep learning is an amazing technology, but definitely not enough to solve AI not by a long shot. These are pretty strong indictments against deep learning by itself. so, what is the third wave the third wave is basically building something that works more the way our brain or mind works. It's a cognitive architecture where the components of cognition are all represented and working together in concert. so, that reasoning and memory and learning and passing and all of those work together work against your long-term memory or short-term memory and your context and this is what I figured out some 15 plus years ago. And this is a technology. We've developed commercialized and perfected to achieve that in a highly integrated way working against a very high-performance Knowledge Graph. so, this is what allows us to give much deeper understanding to have the short-term memory One-Shot learning zero short learning to have reasoning and to be able to handle real conversations. Now natural language is really hard as any you know, I think pretty much all of you will know I don't have time to go through a lot of the problems here, but let me just use one example. Of just how far away we are with chatbots. I can have a conversation with a five-year-old child was just six words. My sister's cat's Park is pregnant and a five-year-old child will have no difficulty in understanding Peter speaking. I have a sister. My sister has a cat the cat's name was Spock. You might think it's male you here. It's pregnant now, you know, it's female and we take that for granted current chatbot technology couldn't come anywhere near that. That as you know, its General ability to learn and understand but this really what we need. so, the difference is really like chalk and cheese the current approaches of chatbots. They are trained offline. And then the model is deployed essentially as a read-only model. Training data is very large. Whereas with third approach you working with an ontology you teach the system General skills and knowledge that are used for all applications. No reasoning and al so, they are black boxes. so, when something goes wrong your remedy is really to throw more data at it and hope that it you don't have catastrophic for getting with cognitive architecture. You can actually zero in on what the issue is and you can fix it more reliably. I just want to talk about with the approach of having a cognitive core that has a lot of cognitive ability that has memory and reasoning and knowledge base and all of that. You can create all of us applications basically by adding the application specific layer that you need you start off with a core strong cognitive cause that already knows. People and places and how to hold a conversation and then for every specific application, you will add the application layer the blue layer here with potentially with additional ontology 's and apis to the backend system Green Layer represent what each individual user teaches the system of what the system learns from each individual user. I think that's all I have time for. so, I'll wrap it up here. Thank you, one question. Thank you very much. one question so, it's really cool because it seems like you're far out ahead of where most of the industry. If not all them are. What are you struggling to tackle right now what problems seems unsolvable to you guys right now? so, we a long way from han-level AI normally if I had to kind of put it on the scale, I would say, you know, the best chatbots. I mean if you put them IQ 10 on and I mean IQ is not the right measure. We may be at 25 with all the additional cognitive ability long way from to han. so, common sense knowledge and reasoning having knowledge about the world is really hard to capture in a system, especially in a system. That isn't on the isn't a robot and of course involving robotics makes the AI problem even harder having to deal with robots. so, it's really the common sense knowledge and reasoning that the biggest problem and of course, we're a very small team. We are only 12 people in the company right now. We just gone commercial with the second generation of our technology now, so, you know, it's not going to be solved by a dozen people. Thank you very much Round of Applause.