IBM maps out Watson’s road to cognitive ECM
The Watson brand has exerted a heavy gravitational pull on IBM’s promotion across a wide range of its products and services for some time now. At this week’s World of Watson event in Las Vegas IBM showed the most compelling presentation of the story I’ve seen so far of how the company’s machine learning services are being woven into the fabric of its offerings.
A critical mass of support (through platform availability, specialised datasets, and domain knowledge born of strategic acquisitions) has finally built up around Watson. We’re still just starting to walk in the foothills of mass applicability and adoption, of course. But that ‘road to cognitive’ is now much better paved than it was before, and more of IBM’s story is starting to hang together.
IBM has also picked up on some of the ways that cognitive services can have significant impacts on familiar technology capabilities. In ECM, for instance, they can help identify, describe, and therefore illuminate ‘dark content’ (say, rich media like video, images, audio which would otherwise languish with under-effective metadata, meaning that it would often fail to get discovered and realise its business value); they can help serve up the most relevant content in context at appropriate points in a workflow; and they can be used to train a learning models about content and business process contexts so that a system continually improves its effectiveness over time.
In IoT, IBM highlights how vast streams of future unstructured data from instrumented devices will be ripe for a measure of cognitive assistance to help spot patterns and take action at the edge (i.e. on the device itself – as I covered in my blog about IBM and Cisco’s IoT tie-up earlier this year), or again at a point of ingestion into a database or unstructured content management system – providing accurate metadata which can be used to inform workflow triggers, or render the information far more discoverable (for analysis and presentation) than it would otherwise be.
We’re still a long way from being able to point Watson at the Internet and telling it to go and learn everything so it can do anything (phew!). And despite a few claims from conference stages of ‘creativity’, to my mind the examples at World of Watson were still heavily human-mediated. We’re talking about “Artificial Specific Intelligence”… not general, all-knowing, all-understanding AI. Watson needs to learn, and that learning can still only practically take place in bounded contexts. So Watson solutions for specific industry use cases (where datasets and domain knowledge are more easily definable) feature heavily in IBM’s gameplan.
As for the ECM space in particular, look out for cognitive-powered applications (from IBM and platform-friendly partners alike) that can bring vastly more (and varied) content to bear in far more useful ways.
When faced with vast quantities of unstructured content (video, photos, etc.), there’s no way to sensibly scale the metadata generation process without machine learning algorithms handling what it’s no longer practical to rely on people to do. (And without accurate and rich metadata, it’s very difficult to really get unstructured content to work for you.) There’s so much video streaming in from instrumented security cameras, for example, that you need systems to watch it for you (and only cognitive ones can watch it like you). We’re at the start of a Big Content cognitive revolution, and Big Blue’s aiming to beat its rivals to being the biggest brain on the block.
Now, taking a step back for a moment: IBM is starting to refine how it talks about Watson, but it needs to go further. It’s very proud of when something called Watson beat the Jeopardy gameshow champions in 2011. Very proud indeed. And yes, it does anchor the Watson story to something that has broad appeal, in at least the US psyche. However Watson the Jeopardy champion-beating machine is only really a part of today’s story in that it acted as a testbed and inspiration for today’s offerings. There’s no big tin Watson box with a logo on it sitting in a bunker somewhere that you can go and stroke; and similarly, there’s no longer a single Watson service in the cloud. There’s a whole family of Watson services, and more.
IBM’s own cognitively-enhanced services for industry are billed as “IBM Watson x” – so you get IBM Watson Marketing, IBM Watson Financial Services, IBM Watson Education, IBM Watson Supply Chain, IBM Watson Talent, IBM Watson Work, and so on. Where it’s a partner infusing their own offering with Watson as part of a wider cognitive ecosystem, it’s “x powered by Watson” – e.g. .WayBlazer Powered by Watson, CaféWell Concierge Powered by Watson, etc. (Keep up!)
With the repeated hammering home of the Jeopardy Origin Story, IBM does risk perpetuating an idea of Watson-the-singular-brain that’s long out of date (and unhelpful). Having made a good job of selling Watson the cognitive concept to customers and partners, IBM will need to work on its story a bit more to explain how that translates into the newly extended Watson family (and its wider partner gene pool). Especially getting beyond the sorts of promo videos that tickle the fancy of futurists.
Concrete use case definition will only come once IBM and its partners can draw on real industry project experience. Expect Watson to have less gaps in its resumé come this time next year once IBM has more adoption stories to share. Just as Watson learns how to do its job better, the more of that job it does… so the company will learn how to promote Watson better, the more Watson it sells. How very… ‘cognitive’.