IBM i and Watson
Steve Will discusses how IBM Watson and IBM i are related.
Recently a LinkedIn group member asked the following question:
“Just curious, I have been hearing lot more about IBM Watson. Are IBM i and IBM Watson related in some way? I know basic things about IBM Watson—that it is a supercomputer and deals with AI as a question answering machine.”
I answered the question briefly on LinkedIn, but I also decided it would be a good topic for this blog. So today, I’ll discuss IBM Watson and IBM i.
First of all, let me set your expectations. I am not an expert on all things “Watson.” Like many people, I watched when IBM’s Watson was a contestant on the “Jeopardy!” quiz show, and was suitably impressed when the IBM-created combination of cognitive technologies was able to triumph over some of the best minds in the show’s history after a long period of learning how to interpret the clues, search for the answers and strategically select from the options the game presents. At that point, IBM had demonstrated that the capabilities of software—running on IBM Power Systems, just as IBM i does—were extending into what people have recognized as “Artificial Intelligence.” So, yes, even from a limited knowledge of Watson, most of what the person asking the question described is true: Watson did start out as a “question answering ‘machine’” and even in that early form, it certainly is “related” to IBM i insofar as each is an operating environment which runs on Power Systems. That much, we all know.
However, in the years since its run on “Jeopardy!” Watson has become much, much more than a question answering machine (though it is still that, in part).
And that’s where my expectation setting must be restated, and I will encourage you to read more from the experts who work on IBM Watson as part of their day-to-day jobs at IBM. I, like you, learn by hearing from them, and there are many excellent articles, as well as descriptions of cases where Watson is making a difference in many varied fields, including medicine.
However, those of us who are interested in the future of applications, particularly business applications, do not need to be experts in the inner workings of Watson to understand, and apply, a few lessons which Watson has already taught us. And that’s the real point of today’s blog.
Lesson One: Watson, and Cognitive, are more than one thing.
Today, if you want to use IBM Watson, there are several ways you can do it. A company or institution can, for example, buy some systems and engage with IBM to create a “Watson” which serves their business. Hospitals and medical research institutions, for example, are using Watson to “ingest” medical research so that it can be used to advise medical professionals or suggest advances in research paths. What does that mean? Well, a human can only read so fast, but Watson can read much, much faster. It can “ingest” (read, comprehend, and make connections among) hundreds of research articles in the same time it takes a human to do a few. This does not make Watson a doctor, but it does make it possible for Watson to find potential relationships in vast amounts of data that a person might never see. Or, perhaps more to the point, Watson can be used to validate (or refute) a new theory about potential relationships because of the large amount of data it can consume.
That’s amazing. And some of the people working in IBM are working on exactly this sort of use for Watson. In fact, to bring it “home” a bit, we have a large group of developers here in Rochester, Minnesota doing exactly that, working with specific institutions to apply Watson to medical applications. Oh, and some of those people have an IBM i heritage, so there is that additional “tie in” to IBM i, even if their Watson code doesn’t run on IBM i itself. (On a personal note, my brother Eric, who was once the Business Architect for Db2 for i, has worked on this part of Watson for several years. He and I have pretty cool jobs. Just sayin’.)
To do this, Watson has had to “learn” to read unstructured information, like the text written in scientific journal articles, assign meaning to it, and draw correlations. This same ability to “ingest” unstructured information is being used in other types of applications, such as “Help Desk” automated “bots.” Initially, they’re seeded with answers to frequently asked questions, but over time they learn the many and various ways users might ask questions and ultimately suggest answers that might not occur to a human who has the same amount of experience.
So, this ability to work in a specific installation on a specific problem set, really means that Watson is many things, and can be many more.
But while that lesson is very, very cool, it’s not the lesson which is most applicable to the IBM i community. No, there are two other lessons, which together make the connection to IBM i much more immediate. They’re the reasons you’ve been hearing more about Watson in IBM i circles. Those two lessons are:
Lesson 2: Watson is also a set of services you can use over the internet.
Lesson 3: IBM i already has the technology you need to use Watson services from within your application!
Another way IBM is making use of the Watson cognitive technology is to make it possible for application developers to use the various components of the Watson technology to do specific “cognitive” tasks. IBM does this by providing a large set of callable services, and those services are implemented in standard ways which exist as options for IBM i programmers.
Examples? You want examples? Well, the very first examples I saw of IBM i working with Watson, or technologies using social media data (another capability of Watson), didn’t come from IBM. No, at about the same time, I encountered both Paul Touhy’s “language translation written in RPG” and Rainer Ross’s use of non-traditional data in his hotel search software.
Shortly after we saw these members of our IBM i community taking advantage of the kinds of Watson services IBM is offering, Alison Butterill asked some of the members of the IBM i development team to investigate what else clients could do with it. She reached out to members of the IBM i community in IBM to ask if they knew of any customers using IBM i and Watson together.
Before long, a set of examples was created. At the COMMON Annual Conference in Orlando this year, the IBM i team showed those examples to packed rooms. And the same happened in Brussels at the COMMON Europe Congress. What did they show?
Well, people on the team are in the process of writing up descriptions of most of them for developerWorks articles and similar avenues of publication, but let me whet your appetite with a few highlights.
Let’s suppose you want to get some information from Watson, working in SQL. Well, Scott Forstie used HTTP Functions and JSON_TABLE to consume Watson web services from Db2 for i via SQL. Through the summer, he has shown how simple it is to use the Language Translator, access the Weather Channel’s almanac and forecast services, and most recently has used Watson’s Tone Analyzer to identify emotions within text. Scott also paved the way for Db2 for i clients to easily move data from IBM i directly into Bluemix and Watson Analytics. This addition of Db2 for i as a data source has enabled IBM i clients to leverage Watson Analytics for analytics, predictive and cognitive computing. While in Orlando, Kevin Adler used Watson’s speech-to-text, text-to-speech and conversation services to create a “Hogwarts Sorting Hat” program, written in Python. And Jesse Gorzinski told the crowd how he asked Watson to do a personality analysis of one of the other well-known members of the IBM i development team, using the Python 3.4 that is delivered in 5733-OPS.
Watson provides some “Analytics as a Service” capabilities as well, and IBM Japan used the combination of Db2 WebQuery and Watson Analytics to provide cognitive predictions for a business in Japan, who are using it to get products to the right stores, in the right quantities, by factoring in things like weather forecasts!
All of this is possible because Watson uses modern, standard services definitions, which can be used by ILE programs on IBM i. Watson publishes easy examples for many of them using modern programming languages, such as python, node.js and others which are now part of IBM i.
The Driveway
To help developers get jump-started in thinking about how to enhance their applications with new cognitive capabilities, IBM and COMMON partnered to create an event, called the IBM i Driveway to Watson. This two-day seminar starts tomorrow, Sept. 6, and is designed to teach the participants about many of these enabling technologies, and to make it easy for them to take the next steps. I can’t say for sure if this event will be repeated, but considering that it reached its maximum enrollment over a week before the event, I think there will be demand.
So if you, like the curious IBM i user on LinkedIn, were a little curious why Watson and IBM i are being discussed together lately, now I hope you have at least a bit of an idea. This blog is far from comprehensive, but that’s hardly surprising. We’re at the very early stages of this new Cognitive Era. How it will progress from here to its more developed stage is still an adventure we all have to take, but what should be clear is that IBM i developers can take full advantage of it.