IBM's Watson: Bringing The Future Together

| About: International Business (IBM)
This article is now exclusive for PRO subscribers.


I recently had the chance to “interview” IBM's Watson.

I still don't know Watson's inner workings, of course, but I have a much better idea of what Watson can do.

And one key takeaway is that it can bring together a host of IBM's services.

Image Source: IBM

In my first article about IBM's (NYSE:IBM) Watson I discussed why this technology isn't a fancy computer language. In this article, I'm going to talk about what Watson is and, more important, one of the key things it does for IBM. Basically, Mr. Watson spans IBM's technology landscape.

So what is Watson
At one point, Watson was really a thing. Think back to the computer playing, and winning, the Jeopardy! game show. At that point Watson was basically a single computer system taking on the specific task of digesting questions, searching through its database for possible answers, assigning probabilities to those answers, and presenting the most likely answer as its response. Add in the complexity of playing against humans and time constraints to the mix, and it was an impressive show of what Watson was capable of.

But that was then, this is now. Today Watson isn't a single entity. When I went to visit IBM's Watson site in New York City, Watson was described as having different iterations. In other words, every customer using Watson gets their own Watson. The takeaway for investors is two fold. First, Watson isn't one thing, using a single name is just an easy way to describe this business unit. Two, Watson can be "sold" many times over. This may sound pretty basic, but it's important to the understanding of what IBM is trying to do with Watson.

Image Source: IBM Case in point, when Watson was playing game shows, it was focused on doing just one thing. Today, Watson does many more things... over a dozen things in fact. For example, digesting and understanding text count as two services that Watson can provide, a third is providing photo recognition and labeling functions. Those are just three of many, but they intertwine (I'll get to that in a moment). The real point here is that customers can get a "Watson" of their own to do any number of different things or multiple things. You can see a full list of Watson's services at IBM's website.

Understanding that Watson does many things helps in understanding one more, and very important, fact. Watson is a key way to connect IBM's strategic imperatives, which include Data and Analytics, Cloud, Mobile, Social and Security. These are the business segments that IBM has been refocusing around because it believes they represent the future of computing today. Notice Watson isn't listed there. Sure, it could probably fall under Data and Analytics, but that would be shortchanging what Watson can do.

Just ask Ted
For example, the Ted Talks series uses Watson to digest transcripts of its speeches and categorize them. Watson does this by understanding the context of what it's read. It doesn't provide absolute answers, however, it provides a percentage to indicate how sure it is that its answer is correct. So, a speech will get multiple categories and each will have a percentage next to it. That allows users to search through and find the topics that interest them without assigning a single variable that may not fully capture the scope of a speech.

But, as noted, Watson also digests photos and video. Video, after all, is basically just a series of photographs. In this case, Watson will examine a photo and try to figure basic things out, for this example I'll just discuss gender and age range. The photo example I was shown broke down gender into male and female (as you would hopefully expect) and age into over 45 and under 45 (I guess 45 is when age starts to creep up on us enough that we start to look "old").

Now imagine that you remembered seeing a Ted Talk given by an older woman about memory. That's all you have, because you're over 45 and your memory has gone the way your looks have - down hill. (That's a joke. I'm sure you look just as good as you did when you were 17 - I know I do... Honest!). The thing is your addled mind recalls that it was a good talk and you really want to show it to a friend (or just watch it again, to refresh your memory). There are a lot of Ted Talks - how do you find it?

Image Source: Reuben Gregg Brewer You could look at still shots from all of the Ted Talks videos and hope you recognize the woman again. You could read all of the Ted Talk titles and hope something rings a bell or suggests a talk is about memory. You could try Google, but that search is probably going to turn up a long list of stuff that has virtually no relevance. For example, just by typing Ted Talks you will likely get every Ted Talk video in your results, even if you put in additional qualifiers like memory - each qualifier you add will bring in more material that has no relevance even if it trims out material from the Ted Talks series. You might come up with other options, but no matter what you try, it's likely to be time consuming.

If you used Watson, you could ask to see all of the Ted videos that were about memory given by women over the age of 45. Watson would then, fairly quickly, provide a short list for you to review. Remember, too, that probabilities are assigned. So you could cull the list further, by weeding out any video that included memory but that Watson only assigned a low probability of the talk being about memory. With the ability to combine Watson's "skills" into one service, your search for a speech just got a lot easier.

She's probably a woman
But Watson is not perfect and Watson knows it. That's why probabilities are so important. For example, in one demonstration of the photo recognition system shown to me we looked for photos of women under the age of 45. On one side of the results were photos and on the other Watson provided a listing of categories under the heading of woman that included man. I thought that was interesting, I'm looking for photos of women and Watson gives back men. So we clicked on the male category.

Essentially Watson's result was saying that more than one person was in the photo and one of the people was a man. But then we clicked on one of the photos and Watson provided all of the categories specific to that single photo... and their probabilities. Man came in with 100%. Woman had a 99% probability assigned to it.

The photo was, indeed, a woman. But the interesting thing was that she was in the middle of a dance move with a somewhat contorted face. Add in fancy dance hair and a heavy dose of makeup, and I can see where Watson would hedge, just a little bit. (I'd love to see how Watson handles Saturday Night Live's famous Androgynous Pat character.)

The point is that you and I might say woman, but Watson isn't human and it provides facts for humans to digest. Based on its estimation, the dancer was almost certainly a woman, but there was a slight chance that the person was a man. Gender is something of a yes/no answer, but other things aren't. Like the context of the memory speech noted above, where percentages have much more meaning for the human user.

Two (or more) in one
Now step back and look at this example again and compare it to IBM's core imperatives of Data and Analytics, Cloud, Mobile, Social and Security. Clearly, there was a component of data and analytics in there. But the data being examined likely lived in the cloud, hosted either by IBM or some other provider. Meanwhile, the search tool is probably meant to be used on mobile devices, which means it has to adjust the way it presents information based on the device being used to access it. Moreover, it wouldn't be a stretch to suggest that the tool would need to be secure and provide the ability for users to share their findings with others in the social spaces they prefer to occupy.

So what seems like a "simple" video search really turns into so much more. And you can see how Watson reaches across the imperatives to provide a single unified service. This example, however, is pretty simple, I was shown a much more compelling medical example which I'll discuss in part III, where I'll talk about how IBM makes money with Watson.

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.