Watson to the Next Level: Alison Butterill and Jesse Gorzinski Discuss the Wizardry of Cognitive Solutions

Cognitive Computing

Paul Tuohy: Hi everybody and welcome to another iTalk with Tuohy. So I'm here in beautiful Brussels at the Common Europe conference. I'm delighted to be joined today by Alison Butterill, the wizard of IBM i, and Jesse Gorzinski, the wizard of open source on IBM i. We will explain the wizard bit a little bit later on if anybody isn't aware of it. So Alison, I'm going to start with you. There's something that's been kind of bothering me late―recently. Have we just gone through a name change? I keep seeing this word "cognitive" all over the place.

Alison Butterill: Well not really a name change. Our business unit that we've worked for for the last little while was called Power Systems, and Power Systems, the physical boxes, still exists. That's what we sell: Power Systems, Power servers. But our business unit has grown and it includes Power Systems but also a lot of cognitive solutions. As a result, the business unit has changed its name to Cognitive Systems, so it includes things like Power artificial intelligence as well as a lot of the linkages to machine learning, deep learning, and other software that goes with it. But we still sell Power servers and for a lot of clients, a lot of our partners, we're still Power Systems.

Paul: Okay so now you've thrown a number of terms and expressions in there so just so we are all clear. It is still Power. Okay. We're still IBM i.

Alison: We're still IBM i on Power servers.

Paul: Okay.

Alison: Absolutely correct.

Paul: Okay. Okay, perfect. Okay so this whole thing of cognitive, all of those things you just said. Okay. Did you want to explain a little bit about what cognitive is?

Alison: I guess―well I'll take a first pass at it and then I'm going to let Jesse add some comments as well, but cognitive is a very different way of thinking than normal IT people have thought when processing data and looking for answers. Cognitive is so much more. It's sort of applying the logic of human brains, human experience to data and figuring out how to analyze, how to work with that data. So it's a lot of different aspects. There is natural language support. There is inference support―drawing inferences from seemingly unrelated pieces of information. There's also extending the experience and doing research into nontraditional data forms, things like social media and so on. Cognitive is really interpreting data in a different way, something we haven't done before.

Paul: Okay.

Alison: Jesse, do you want to add a couple?

Jesse Gorzinski: No. The whole cognitive technology is really―in some cases we break it up into what we say the URL. Right? We have this technology now that can understand―so that's the U―understand information in much the same way that humans can themselves. So rather than traditional computing that would take for instance a string of text and pick out maybe key words, identify words that would be synonyms to those words and apply traditional algorithms. We actually have the ability now to understand what that text is actually saying, which is the natural language support that Alison mentioned. From there, we get to the R: the reason. So Watson actually has the ability to reason and draw conclusions from that data that are based on much the same criteria that the human brain would be using to base those decisions on. So it's very, very powerful technology. Then that L―that last bit of the cognitive URL as we sometimes say―is the ability to learn. We've been talking a lot lately with customers at events about deep learning and machine learning and the ability to have models that react and learn, as new information is perceived, as new information comes in and also can learn from itself when results aren't necessarily optimal. So those things really differentiate cognitive from the traditional computing models.

Paul: Okay so and again, if I'm oversimplifying here, please say so. So really when we're saying cognitive, are we just talking about Watson? Is that really what we've talking about?

Alison: Well I think Watson is a good first step, a first look, and an easy way to get started, but cognitive is more than just Watson. Cognitive is also things like artificial intelligence―and there are other products that we have that support artificial intelligence, machine learning and deep learning―but Watson is sort of the implementation in the cloud if you will of how to work with some of those concepts.

Jesse: Yeah, exactly. I don't really have much to add onto that, but Watson of course is the product that we are offering in the cognitive space that we're hoping people can latch onto. We're hoping that we've delivered something that the world can use to not only improve business, improve business intelligence, but also just to make the world a better place. And Watson is the offering that is on the forefront of cognitive computing.

Paul: Yeah and just for people listening in, I mean, if you haven't really been following what has been going on with Watson, it is making incredible differences in areas. I know in the cancer research areas and all of that that they're doing those great things, which is what I always associated Watson with. I would read about things and say "This is great. It's marvelous what they're doing." Then I see this cognitive thing―so how does this tie in with IBM i then? I mean, there are not many IBM i shops who are doing cancer research here, that type of thing. So how does i fit into all of this?

Jesse: Well one thing that IBM i shops typically do have is a lot of data, right, and you pair that with some of the capabilities that cognitive machines can now do, things like identifying patterns within that data at a large scale, right. Humans have a great ability to look at a sequence of numbers, a sequence of paragraphs from people identifying mood changes and pattern changes and things along those lines, but we can't do it the large scale, right. If you have millions and millions of patient records, for instance, you're talking about cancer. Humans can't identify patterns in the same ways that people can. So that capability translates well to i customers and many other industries whether you're doing logistics, banking, or finance, the ability to among other things identify trends, identify possibilities for business growth within that data by understanding―again it's understanding, reasoning, and learning―by understanding that data and reasoning from it. That's how it fits with IBM i because IBM i is a platform that right now is―it's always been a platform built on data. We have a lot of customers with a lot of data, with a lot of potential within that data and so cognitive is what can bring IBM i to an even further place in the advancement of business.

Paul: Okay so we're talking here about sort of like if I've got these masses of information in my databases and some way of like giving that data to Watson, asking it do analysis, this cognitive analysis. In other words my nontraditional simple SQL select statements type of thing―

Alison: Correct.

Paul: Of like how many are there? How much does this come to? Of actually doing this yes, cognitive analysis. I now understand the word. It's brilliant. [Laughs] I'm feeling so much smarter. But doing this sort of cognitive analysis and getting that information back so that being the case, how? I mean it was fine saying it does it, but how? How do we do it?

Jesse: So in the case that you're talking about, you have a set of data, and it resides on IBM i or possibly on some other data source that you have, another database, whatever. You can take that data and you can send it Watson through this interface called Watson Analytics. Watson Analytics is kind of what I call the nonprogrammer path, meaning you don't have to be a programmer to use it. If you are somebody who has data, you know there is value in this data somewhere, you do exactly what you say: You take what you have. You can give it Watson through this interface called Watson Analytics. There is varying levels of free to non-free that you can use and then you can ask Watson questions. Some of the important aspects of Watson Analytics are that it incorporates some of the capabilities that we've already talked about, the ability to ask questions for interest―for instance using natural language and having Watson just understand the question, understand how to render answers in a way that makes sense. That's the primary way that you do it if you're just interested in taking the data you have and getting insights from it. That's a very short, easy path to do so.

Paul: So is this like a very smart Siri that we're talking about? [Laughs] I mean in relation to my data, my information.

Jesse: So Siri―Alison has some thoughts.

Alison: No, I was going to say, well in some ways it is similar. You speak to Siri and Siri has the power of all of the applications you might have on your phone and whatever you're connected to to pull on information so you could ask Siri for directions to some place and Siri will do the search of the web, find the location, and draw out the map for you, and tell you how long it is, and how long―and will track your progress using GPS. So in some ways Siri is acting like a part of what Watson might be able to do as well. Again, Watson is a lot more than that, but that is something you could―

Paul: That's what I meant. It is really, really smart.

Alison: Yes, exactly. Really, really smart. Yeah.

Paul: Right.

Alison: PhD. [Laughs]

Paul: So the mechanics for this? Do we need special software on i? Do I need to buy a Watson product or something to do all of this?

Alison: Well actually I―just―

Paul: Oh, I'm sorry.

Alison: I was―that's okay. Jesse talked a lot about using the data and the large amounts that data that IBMers―I'm sorry―IBM i companies have on their machines, but another aspect of Watson that we also want to talk a little bit about is the ability of taking your existing application and using some of the capabilities that Watson provides by hooking into those applications. So for example Watson has a number of services that it offers: language translation, personality analysis as well as other things, the ability to search nontraditional data. Looking for tweets for example: examining tweets, searching for particular things inside tweets. These are things that our IBM i clients in their applications could be talking to Watson to get added value so it is not―while data is the one that is the most obvious use for many IT people, a less obvious but maybe even more powerful use is what I can do now that I could never do before―

Paul: Yeah.

Alison: So I just wanted to make that point before we moved onto the pieces.

Jesse: Exactly.

Paul: Well that was my little delve into the Watson thing. As you know with the language translator, which was great fun, and so marvelously easy to use.

Jesse: And it wasn't millions of line of code.

Paul: Oh no. Two. Two lines of code for the Watson bit.

Jesse: Right.

Paul: Yeah. I had more lines of code for the screen handling than for Watson. [Laughs] All this is great but again, there is no special software that I need, anything like that?

Jesse: No, you know I'll talk about the plumbing for just a minute―

Paul: Yes, please.

Jesse: If you talk to Watson, you're using plain HTTPS, rest-based APIs. So it's industry standard web services API calls the EUS to talk to Watson, which means, to answer your question, no you don't need any special tools, you don't need any special languages. Virtually any language can do it because any language can use HTTP, HTTPS-based restful API calls―with varying levels of difficulty, perhaps, or integration. But virtually any language can do this stuff. So from IBM i, you demonstrated RPG is one very easy way to do it. Straight up SQL, there are people using that as well. We have open source languages―i.e. PHP, Ruby, Node.js, Python. We have ILE languages―C, C++, etc―and you can talk to Watson from any of these. You don't need any extra tools and you get coexistence of the application with your data and the ability to tie it in with the cognitive system to just kind of close that loop.

Alison: The one thing is you do have to be at 7.1 to get those HTTP services―correct, Jesse?

Jesse: Correct.

Paul: Yeah. [Laughs]

Jesse: For the SQL services, you do need to be at 7.1 or newer. Good point.

Alison: Yeah.

Jesse: Also the same applies for Python and Node.js.―

Alison: Correct.

Jesse: 7.1 and newer.

Paul: So where do people go then to find out about all of these Watson things that they can do? Where can they go to get the information on it?

Jesse: There is a place on the web called the Watson Developer Cloud. Do we have the ability to embed links in the transcript? We probably do.

Paul: Yeah, we will indeed.

Jesse: So there's a place called the Watson Developer Cloud that has descriptions of all the varying capabilities. There are over 50 capabilities that are uniquely defined within Watson. It has information and resources about all of those as well as links to what we call the API Explorer, which lets any programmer for any language figure out how to call these things as well as working code samples, right. So in one of my sessions at Common America this year, in 3 minutes I demonstrated how to pull down code from Watson Developer Cloud, be running Python and talking to Watson in a matter of, I think, it took me 2-3 minutes right so it's very easy to get going. Watson Developer Cloud is definitely a great first place to start.

Paul: Yeah. I must―I mean the little bit I've played with it, it is so easy it's embarrassing, you know, considering what you're getting from it. Okay so listen we're nearly finished here but before we go because as I'm looking across the room here, I'm looking at a sorting hat. Okay, so this ties back on the wizard things so come on. What is it with the Harry Potter and the sorting hat and the wizards of and the cloaks that you guys have been wearing here at the conference? [Laughs]

Alison: Well I think I'll take the lead on that and let Jesse give you his final opinions. But last summer Common North America was sharing with me that their theme for the conference this year was Harry Potter. They actually held their Common North America event at Universal Studios, and they had rented out Harry Potter World for the evening event. And my brain got going a little bit―and so did Google, with my fingers on the keyboard of course―and I found out that there was an IBMer who had built a sorting hat. My head got spinning, thinking as we moved into the world of Watson, what a great way to have a little fun and to show some of the capabilities of Watson. So in my head, I thought this would a very simple project to build a sorting hat and tie Watson personality analysis to it so I went to Jesse and asked Jesse to help, to see if Jesse and his team would be able to do that. So I handed it over to him and I'll let you tell him his side of the experience. [Laughs]

Jesse: Yeah, so Alison came to me with this idea and said, "hey, you know, I'd like to build a sorting hat." And thankfully she had already actually found another IBMer from the West Coast, I believe, who had already built a sorting hat and documented some of his experiences using, I would say―he used different technologies on the Watson side, but very similar. You know I said "yeah, this kind of suites that bill of what my team is good at. There's open source involved, there's emerging, kind of high tech cool stuff involved." I always brag to my colleagues that my team has the most fun at IBM, so I said this is yet another way to prove that to be true. But it was very fun: the components of the hat, we got a Raspberry Pi, which is something that, you know, 8-year-olds are playing with these days. It's very cool technology. We wrote some code in Python, and then we rewrote the code a few times, but we had fun doing it. People always hear me talk about how Python is fun. But yeah, so we used Raspberry Pi. We run Python on it and Python has very simple as you mentioned services calls to IBM Watson so in our Bluemix account, we actually tailored a conversation. We used the Watson Conversation API as well as text to speech, speech to text and some other things. So we tailored a conversation around nothing more than choosing a house for a person based on that person's personality attributes. That's what we kind of showboated and we take that to events and things. We actually also built a conversation service that we haven't been demoing that let's you learn about magic and spells and things. So you can tell it stuff that you like to learn about at Hogwarts and "gee, I'd like to find a spell to turn my friend into a frog," and it would actually teach you some of these magic things. In all, you know, it was really fun. It was a very fun project. I'm thankful Alison had the idea and also thought of my team as a great fit for making that happen.

Alison: Okay but now I want the spell that turns people into frogs. [Laughs]

Paul: And I think that's good note to finish on. So Alison and Jesse, thank you very much for taking the time to talk to me. I'm looking forward to the next occasion that I get to see the sorting hat. Right now I think the three of us have earned one of these fine beers I believe that you can get in Brussels.

Alison: I've heard that.

Jesse: Absolutely.

Alison: I've heard it's really good.

Paul: Okay. So that's it for this week, guys. We're going for a beer. I suggest you go and do the same.

Jesse: Thank you, Paul.

Paul: Tune into the next one guys.

Alison: Thanks, Paul.

Paul: Bye for now.

Paul Tuohy has specialized in application development and training on IBM midrange systems for more than 20 years.

Like what you just read? To receive technical tips and articles directly in your inbox twice per month, sign up for the EXTRA e-newsletter here.

comments powered by Disqus



2018 Solutions Edition

A Comprehensive Online Buyer's Guide to Solutions, Services and Education.


Paul Tuohy on IBM i Highlights of 2016


Paul Tuohy Recaps 2017

IBM Systems Magazine Subscribe Box Read Now Link Subscribe Now Link iPad App Google Play Store
IBMi News Sign Up Today! Past News Letters