One of IBM's (NYSE:IBM) big talking points over the last couple of years has been Watson, the nickname for its cognitive computing system. Although this is just one offering among many the company provides, it is a vital link between many of IBM's so-called "strategic imperatives." After reading two conflicting reviews of the technology here at Seeking Alpha, I figured it was time to go to the horse's mouth - I asked IBM to explain Watson to me.
IBM is in the midst of remaking its business around what CEO Virginia Rometty has taken to calling strategic imperatives, including Data and Analytics, Cloud, Mobile, Social and Security. As part of this shift, IBM has been selling off businesses that don't fit with its long-term vision for the technology industry. These have generally been lower margin businesses with limited growth prospects such as those built around commodity hardware items like laptops.
The impact of the transition, however, has been a declining top line. Investors have justifiably been concerned about this, even though the growth in the strategic imperatives has been impressive. As a group they grew 26% in 2015 and now make up just over a third of the top line. There's clearly still a long way to go, but based on the growth IBM has seen within the list of imperatives, it looks like it's shifting into the right spaces at the right time.
IBM has made and remade itself many times in its over 100-year existence, shifting as technology and customer needs change. I expect the current corporate rework to be no different than any of the others, sometimes bumpy along the way but ultimately successful. So I'm an IBM believer, if you will.
Which brings us to Watson. It isn't one of the core imperatives, but it is expected to help drive the company's progress. That's because it brings together many separate parts to make a whole. I look at Watson as a way for IBM to marry the core imperatives.
So What Is Watson?
I had a vague understanding of Watson as I started to research IBM as an investment when the company's yield was in the 4% range. To be honest, you don't need much more than that. However, I was concerned about Watson when I read an article by Dana Blankenhorn that equated it to, basically, a glorified computer language.
This and one other article (taking a more positive view of the technology) led to me to reach out to IBM and ask, "What is Watson?" The company, very generously, put me in contact with Stephen Gold, who heads up marketing for Watson. He and I had a discussion about the technology, but, to be honest, I was still a little fuzzy. To help put images to words, IBM offered to let me "visit" Watson in the company's New York City Watson office.
I put "visit" in quotes because it was more of a demonstration of what Watson can do and, more importantly, an explanation of how it does some of what it does. The best part of the visit was being able to stop the demonstration and ask questions or paraphrase something that was just explained in language that made more sense to me. (Tech guys often talk in an acronym language that only other tech guys understand.)
It's not a language
First and foremost, Watson is not a computer language. That said, language is an important component of what Watson does. Take the word "exposure." It's a fine word, but alone it holds little meaning, just like so many other spoken words, particularly in the English language. What do I mean?
If someone said to you, "Let's check your exposure." What might they be talking about? If the context was insurance, then the discussion is likely about the liability risks you are facing. In healthcare it could be a question of whether or not you've been in contact with a disease. In finance you're probably talking about asset allocation. And in photography you are discussing lighting, shutter speed and lens aperture.
There are multiple steps to figuring out which one of these is the right definition. First, Watson needs to pick apart a sentence, basically dividing it into nouns and verbs and whatnot. This is a key step, and I can see where this bit would look like Watson is about learning how to properly write questions so it can understand what's being asked.
But step back for a moment - just identifying the parts of a sentence doesn't give Watson the ability to properly discuss the statement "Let's check your exposure." What allows Watson to properly answer that question is its ability to discern the context in which the statement was made. In other words, by looking at everything that's being said Watson can determine what you are talking about and then figure out what to do next.
The folks at IBM ran me through a couple of examples of what Watson does. Some were more impressive then others, but one example stuck in my mind because of the language component. The company wouldn't reveal its partner's name, but an insurance company is using Watson to help increase online sales. According to IBM that customer has seen a high single-digit uptick in online sales because of Watson.
First off, that's a huge benefit Watson has provided this customer. That, in turn, shows you just how valuable this technology can be. The second part, however, is the more interesting one technology wise. Watson is "manning" a text chat box where potential customers can ask questions about insurance. Now the context is clearly insurance, but no one has taught the potential customers how to properly communicate with Watson.
They are typing in plain English and Watson is figuring out what they are asking well enough to increase the company's online conversion rate. In fact, some customers even end their chats by writing "thank you" - even though they know it's a computer they are chatting with. In other words, Watson communicates so well that people are personifying it.
Now I noted that context in the above example is known in advance. However, that isn't something that happens overnight. Watson is provided with information, essentially creating a database. The information is usually curated in some way, so the content, as with the insurance example, is germane to the topic with which Watson will be working. But that's only step one.
Step two is testing Watson. That means asking questions and getting answers. But, like a human, Watson doesn't always come up with the right answers at first. Watson makes mistakes while it's learning. It understands things in the wrong way and pulls the wrong answers out of the information it has at its disposal. The team working with Watson then corrects it and tries again with another question. The time this takes depends on a lot of different variables, of course, but one customer took a year to train Watson. The chat bot that Microsoft (NASDAQ:MSFT) embarrassingly had to pull offline recently after users trained it to be a bigot is an example of what can happen if the training isn't done well.
But what is the training that IBM is giving Watson? The company isn't teaching people how to communicate with Watson, which would essentially mean that Watson is a programming language. The company is teaching Watson how to communicate with people properly about a given topic. One part of that was understanding a sentence and the other was understanding exactly what was being discussed - in other words proper context.
Having had the chance to "meet" Watson, I can say confidently that Watson is not a fancy computer language. It's so much more than that, but I'll get into an example of that in the next article. That said, thanks to the generous offer of a Watson "interview," one of my big concerns about the technology has been put to rest.
Disclosure: I am/we are long IBM.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.