Cognitive computing technology is not ready for prime time, but businesses may be able to tap into some of its...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
benefits now -- particularly by approaching content management in a new way.
In early 2011, when IBM announced that Watson, its smart supercomputer, beat two human champions in the Jeopardy Challenge, many of us received the news with a mixture of disbelief and wonder. Yes, with enough time and effort, you can train a machine to find the best matching answer to a given clue.
Since then, IBM has established the Watson Group as a branded business unit and, in January 2014, announced plans to invest $1 billion in efforts to commercialize cognitive computing technologies. The company is also funding related research at major universities around the world as part of its efforts to support innovative computer science disciplines.
It's all about pattern matching
While Watson's ability to reason through complex problems may seem revolutionary, we've seen this play before. Like so many computational quandaries, the right answers will depend on formulating the right questions. And it's a long way between demonstrating research results and translating them into commercial products and services.
As enterprise search experts Susan Feldman and Hadley Reynolds described it, cognitive computing "addresses complex situations that are characterized by ambiguity and uncertainty." They emphasized that "cognitive computing systems make context computable. … [These systems] provide machine-aided serendipity by wading through massive collections of diverse information to find patterns and then to apply those patterns to respond to the needs of the moment."
What's new about cognitive computing is the varied techniques for pattern matching that you can bring to bear to recognize different kinds of digitized information, from predefined entities within structured databases to the bit-patterns of streaming rich media. Bob Kahn, one of the inventors of the Internet, recently encapsulated the key issues: "How do you make a cognitive system happen? You need basic rules of the road where semantics are important and type registries are important. It's also going to take patience to build a community ecosystem and figure out the target problems."
If technologists are excited about the prospects for cognitive computing, industry watchers also need to inject a dose of caution into the mix. Watson remains a work in progress. IBM is partnering with a cross-section of its customers around the world to develop purpose-built reasoning systems. A research symposium in October 2014 featured demonstrations of smart pattern-matching systems applied to oil and gas exploration, as well as genomic research. These are large-scale problems within an enterprise where the potential payoffs justify systems-level investments.
For customers and prospects with shallow pockets and a curiosity about the future, IBM is launching Watson Services for IBM Bluemix. These will be RESTful services, accessible to anyone with a Bluemix account (IBM's branded cloud offering). In October, IBM announced an external beta release of the first eight services in what promises to be a large and growing set: user modeling, language identification, machine translation, concept expansion, message resonance, question and answer, relationship extraction and visualization rendering. Each service performs particular kinds of cognitive functions.
Developers can begin to mash up their own content with these cognitive computing services through the cloud to extract further insights. For instance, the user modeling service characterizes personality traits by analyzing peoples' writing samples or Twitter feeds. This service might be useful for lawyers trying to pick jurors from a jury pool or human resources managers evaluating candidates to hire.
As promising as a services-oriented approach may be, currently it has limitations. IBM trains the various services so they are tuned only to predefined information collections. For instance, at present, the question and answer service just addresses travel and health care information.
In the future, enterprises will need the capability to add and analyze their own information sets -- and extract the cognitive connections among the content components. Tuning these private information sets will take specialized skills and expertise.
How can businesses start using cognitive computing?
Although these services are not yet ready for prime time, what should companies do to profit from the prospects for cognitive computing in terms of Web content management? They need to watch two long-term trends.
First are new approaches to problem solving. Most website owners are familiar with search engine optimization -- organizing their content so their target audiences can find what they want and discover what they need. Website owners must pay attention to their information architecture and how they semantically enrich their content for findability.
There's a virtuous cycle to metadata enrichment and semantic modeling. More connections among data elements lead to greater opportunities for serendipity within user experiences. Website owners need to renew their focus on their information architectures and begin to create advanced models for user interactions.
Second are new approaches to machine learning through metrics. Computers are good at counting, but we need to teach them what to count. Most website owners are implementing Web metrics -- often beginning with Google Analytics -- to track the behaviors of site visitors and to tweak delivery options. You can automate what begins as a manual activity with advanced analytics -- provided you have models for the processes.
This approach to analytical thinking requires that website owners identify their goals and objectives and then specify the metrics that measure the results. Websites generate a lot of data, and it is essential to determine what needs to be measured. This happens through trial and error, iterative improvements and optimizing the feedback loops so machines appear to learn through experience. Investments in big data analytics that measure the business results are a necessary part of the learning cycles.
In sum, we can make context computable by breaking problems down to the point where the implicit becomes explicit. Only then can digital technologies be applied to solve those problems.
Cognitive computing in the real world
How have semantic technologies changed the Web?
Designing WCM architecture for the semantic Web