Stephen Gold came up next to talk about Watson. Watson is transitioning from a proof of concept – a grand challenge project – to a business solution. This is being driven primarily by the rapid expansion in unstructured, uncertain data. Not just a need to process new data sources, also about being able to reason in the face of uncertainty. IBM sees Decision Management and Watson as being the first of a wave of Cognitive Systems that will increasingly replace programmatic computing in the same way that these systems replaced tabulation systems. Time to move from interfaces on a data source to something more useful.
These systems tend to use both structured and unstructured data, apply context to derive probabilistic answers, support discovery not just search, use natural language and support big data. Four key characteristics:
- Designed for statistical analytics
- Scale in
- Automate system and workload management
I prefer to talk about Decision Management Systems that are agile, analytic and adaptive as I think the hardware management pieces have nothing to do with cognitive computing and allowing expertise and policy to be added and rapidly changed is important.
Watson itself is a package of many technologies designed to understand language patterns and speech, generate hypotheses and evaluate them, adapt and learn from the responses to its decisions (it’s adaptive).
After its success on Jeopardy Watson is in increasingly being piloted and deployed as a commercial solution. Initially in Healthcare but with financial services, government and contact centers. The healthcare scenario is about improving advice in a decision support context for medical professionals, an area where there is a lot of unstructured data and lots of data flowing in every month as research continues.
From a technical perspective this means adding 1,000s of users, ingesting new data more rapidly, becoming stateful to remember conversations, adding supporting evidence presentation and allowing longer, less bounded questions. IBM is also bringing Watson to market on the cloud as a service as well as integrating it with its more general Decision
Interesting example from the healthcare delivery world reinforced how important explanation is (a common feature of Decision Management Systems) and how critical organizational change (getting Doctors to use it) will be.
The question for me is not why to use Watson in the future, the question is why not to use a Decision Management System now?