Cognitive Connections Architectures, Use Cases and Code

Editor's Notes

• #2: Hello and good afternoon. This session is Cognitive Connections: Architectures, Uses Cases, and Code. I’m Van Staub … partners … And over the next 45 minutes or so, we’ll be discussing and more importantly showing cognitive and its potential when combined with social data from IBM Connections and Connections Cloud.
• #3: Normally, I would gloss over this. But I want to point out that this session isn’t an IBM product discussion. You’ll hear about internal changes we’re making. And you’ll see code that I’ve written to demonstrate cognitive use cases. So while it’s not entirely in the product today, it is absolutely real, and something you could cetainly do.
• #4: Message: We are moving from the social era and the cognitive era. ~2:00 minute(s) This is going to sound cliché, but I really do think we’re on a journey – clients, business partners, and IBMers alike are being shaped by this coming age of cognitive. Now our roots are in collaboration – most recently social networking. We productized the ”social network” and called it “IBM Connections”. We gave every stakeholder in a business the ability to have a voice and the tools to make that voice heard. But a result of these social and business tools, we have a lot of information - maybe too much information. So I think we’re at an inflection point where we can take all this information – blog articles, forum topics, status updates, the list goes on – and process it, understand it and use it better through cognitive assistance. That’s where we’re headed, and we’ll start taking the steps to get there.
• #5: Message: We are changing – becoming a cognitive company ~ 1:30 minute(s) Today when you think, IBM, what word comes to mind? You might be thinking: hardware, software, maybe services – hopefully no one is thinking “typewriters”. For the past five years or so, IBM – the company - has been dramatically changing. We are becoming two things: cognitive solutions and cloud company. Cognition will be woven into everything we do and cloud is our delivery. But infuse cognitive into everything we do, we need to make a few changes. And that’s where there is a lot of excited people. Our brand is changing [long pause]: changing its internal culture, we’re updating technology, and we’re finally embracing cognitive – you’ll see the monicker Watson Work to indicate that.
• #6: Message: Connections is changing – for the better On our culture, you’ll hear the word “Pink”. We have Connections 6.0 – the next on-prem version of Connections in Blue. And we have Connections Cloud – the Green. And then we have Pink, which I think is what you get if you mixed blue and green. Pink is a transformative initiative led by Jason Gary and the development organization to modernize how we develop and deliver software. To do it at scale and speed. To do it more transparently - with customers, partners, and IBMers alike. To hear Jason pitch it; it’s like hearing a vision from a startup. So why do you care? As a result, we’ll start to see re-investment into APIs, frameworks, and the developer ecosystem. And that’s what we need as we begin to embrace cognitive services with IBM Connections and Connections Cloud.
• #7: Message: Technology has change, and so must we. Let’s take a look at technology. During the OGS, you’ll see winners from a recent cognitive hackathon we did. I was on-site at the Austin Hackaton. And before we got started, I polled the room of 30 developers, designers, and entrepreneurs. I asked which teams were using Java to develop their application. Out of a room of 30, how many hands do you think were raised? Even to my surprise not one. Not one team wanted to use Java. And yet this is the foundation of the IBM platform. It’s an enterprise stack – and mind you a very good one – it’s secure, stable, and scalable. So if they didn’t want to use this, what did they want?
• #8: Message: Technology has change, and so must we. Well it’s this – “emerging” technology. This is the technology that we’re starting to adopt in our products. For starters, it’s not Java but rather Javascript … something that until rather recently was what we used in the Web Browser. Now Javascript along with platforms like Node.JS and open source libraries like Express, we can write entire server applications. And where we once had very structured data, today’s social and cognitive data are not. So we turn to NoSQL databases like MongoDB or IBM’s Cloudant. And on the front end, it’s React and Angular. These new user interface technologies almost effortlessly bridge the backend data with the end user. We now live in this world of an increasing pace of technological change and startup mentality. “Fail fast" , “break things”; these are the mantras of the most successful software companies. When it comes to innovation – rapid prototyping and growing and idea into a business or product – this is what lets me go from idea to something tangible for my clients or partners much more quickly. I make the joke that when was the last time a PowerPoint went viral? Probably never. But an app that you create, that will.
• #9: Every journey needs a guide … I’m a little biased and I think business partners can be those guides. Before I started on this presentation, I spoke to two IBM Business Partners already building cognitive. The first was Opentopic and they’re redefining the Digital Agency with cognitive ... Doing things in marketing campaigns that without AI would not be possible. Oh and they’re also building integrations for Watson Work Services and have a booth in the Expo Center so please seek them out and ask them on their thoughts on where we’re headed and what they’re building to get there. But essextec said something that stuck with me. They said there is big c and little c in cognitive. The little c, is where we can all get started and right now that takes the form of chatbots – these intelligent text agents that help us. As a partner or client, we can do little c use cases today. I think you’ll e surprise with what you can do with a little bit of cognitive. The other is the big c … this is large scale discussion. And it’s something we can wait for … because it takes time to build something truly novel. And that’s where we have the desire in Watson Work, which is to infuse cognitive in everyday business interactions – email, instant message, social networking.
• #10: My little c came after seeing this. I went to Georgia Tech and picked up on this article immediately. A professor in the AI department had written a chatbot –to answer students questions. Think about it, every semester you get a new batch of new students … and what do they do, they ask the same questions. So the professor and a group of TAs create a chatbot called Jill Watson to answer student questions. And the headline is classic … students didn’t know their TA was a compute. I mean these are the smartest computer students in the country. This was so cool, I had to build it myself.
• #12: Let’s build that. We’ll be using IBM Bluemix to deploy custom code as noted in green. We’ll also use a variety of IBM products: Connections Cloud Community forums similar to the forums used by the TAs at Tech. IBM Workspace and Work Services to act as our Q&A chatbot interaction. IBM Retrieve and Rank to cognitively provide the answer to the question asked. To answer any questions we need data. Fortunately, we can cull that data from IBM Connections Cloud forums. We’ll use some custom code that I wrote to do that. The code is a Javascript client that converts Connections data from ATOM into JSON. Once we have data, I use another application I wrote. It’s job is to take that raw data and feed it into Retrieve and Rank. We have to store the content in Retrieve and Rank as well as train Watson so it knows when an answer is a good one. What I’ve done is used to do this for us. Because if you recall a forum has an ability to mark a question as an answer. So we know when a reply is an “accepted” answer. And we also have instances where people “like” a reply. This acts as a way that we can treat some of the replies better than others. Accept answers are the highest. Liked ones are someone in the middle. And just replies are low but still relevant. This process is known as ground truth. Now lastly, we use IBM Workspace to be the UI that students as questions. They type something in and Work Services knows when a question gets asked – that’s cognititve. Once we have this a trigger, custom code goes out to retrieve and rank and asks the question, gets a reply, and if it’s good enough sends it as a reply into the space.