It’s elementary

IBM has been extolling the success of Watson at winning the game show Jeopardy for some years. I must have seen three or four presentations about that since this success but that has amounted to little more than “wow, aren’t we cool”. The company hasn’t really managed to explain who might use this in the real world: until recently. I attended an IBM big data and analytics day and one of the sessions was on Watson. Finally, I get what Watson is about because the company was able to explain how it is productising it and how its early adopters (somewhere between 10 and 20 customers) are using the product to help with real-world business issues.

The first thing to understand about Watson is that it gets better at what it is doing the more it does it: machine learning. But if you think about Jeopardy it was not about Watson getting better at answering the questions it got asked – what it was doing was getting better at playing Jeopardy – that’s a subtle distinction but, I think, an important one. It also demonstrated some advantages to being an essentially dumb machine as opposed to human. For example, one of the categories for Jeopardy questions was “chicks dig me” – the man who chose that subject no doubt thought that the questions would be about girls who liked someone but, as far as Watson is concerned, without further information chicks could just as likely refer to baby chickens as girls. In fact the questions turned out to be about female archaeologists and anthropologists.

Right now there are two Watson products: Decision Advisor and Engagement Advisor. There is a third product that is running in the IBM Labs and there is a fourth planned, though that is probably a few years away. The product names are a giveaway for the functionality that is offered: Decision Advisor collects facts to help you make decisions and Engagement Advisor builds on top of Decision Advisor to help you engage with customers, especially in a self-service environment.

The markets being targeted by IBM are primarily financial services and telecommunications and secondarily retail, technology and healthcare and IBM already has early customers in each of these areas.

One of the questions that always comes up about Watson is whether it replaces people. For example, if Watson can do diagnostics and recommend treatments in healthcare environments then do you need doctors? The clear answer is yes: Watson can’t see the colour of someone’s skin or eyes, it can’t detect body language and it has no psychological capabilities. On the other hand doctors do not typically have the time to keep up with all the latest medical research or treatments so Watson can provide a very significant supplement in that area. Even if we look at Engagement Advisor where a user might be liaising with Watson online to determine the most suitable mortgage or insurance policy – these are largely areas where providers are already looking to automate processes – Watson just lets you do that better.

So, I’m finally impressed with Watson. I can see that it has the potential to offer some real benefits even though IBM is still some way from taking the Watson products beyond its early customer programme.