IBM Watson: a sophisticated data analytics & insight engine
What can the world's most sophisticated television show computer systems contestant do for business? jStart has been busily working with IBM's Watson Solutions Group to put Watson to work?
This has been the question that many have been asking since IBM's Watson system managed to defeat two Jeopardy! champions in the much hyped television event in February of 2011. Which, granted, was impressive. But what can such a powerful system do for businesses; how can it help solve challenges that businesses face, today?
What is Watson, really?
At its heart and soul, so to speak, Watson is a very sophisticated data analytics and insight engine. The system was designed to be able to consult a wide variety of information sources to be able to answer questions for which it may not have any specific knowledge of...think of it as a system which is extraordinarily good at being a research assistant--gathering data from innumerable sources, sifting through mountains of information, and then coming up with relevant answers/suggestions...because, in fact, that's exactly what Watson does. In fact, it does this better than any other computer system that exists on the face of the planet today...and does it in a very quick and timely manner.
Watson leverages numerous existing IBM technologies
How did IBM do this? Well, it wasn't a coincidence that the company has been tackling many of the issues that such a system would require for years:
- Gathering, organizing, and sifting through large amounts of unstructured data
- Using sophisticated text-analytics to understand the context of the items of interest within a broader scope of the text it's discovered in
- Being able to weigh various potential answers to make a call on the best answer/action
- Perform very sophisticated statistical analysis on the information gathered to understand 'relationships' and 'correlations' between data--many of which are not apparent and/or very subtle
- Leveraging powerful back-end systems to store the data, quickly complete very sophisticated calculations, provide answers in an accessible manner...within seconds
By having all of the 'ingredients' at it's disposal, IBM was in the unique situation of being able to pull together the resources and experts needed to construct the world's first computer system that could reliably answer questions based on unstructured, vast, and non-sensical relationships between information--and then provide an answer...very similar to how the human mind creates relationships (which may be subtle and/or non-obvious) and generates insights based on those relationships. Is Watson the realization of the long held computer science goal of artificial intelligence? No. However, with it's understanding of unstructured data, it's ability to relate disparate pieces of "knowledge" together to knit a cohesive picture of what an answer may be, as well as its built-in self-learning systems, Watson is light-years beyond today's expert systems, which is as close as production systems have able to come to. While expert systems are highly "intelligent" in a very narrow niche of expertise, they can not answer questions which move outside of that area of expertise...while Watson--as demonstrated on Jeopardy!--has the potential of doing.
How Watson Works
Not exactly an easy question to answer, but at a high level, think of Watson as a series of components, each working together to create the sophisticated system that it is:
- DeepQA, a system that embodies Watson's capability of the sophisticated analysis required to come up with its "best guesses", DeepQA leverages IBM natural language processing technology to "understand" the context of the question being asked, as well as provide a degree of representation and reasoning based on the processing of the information.
- Hardware. And lots of it. Watson leverages top of the line IBM hardware technologies, including storage, processing, and massively parallel systems. Realizing that not everyone wants to make such a significant investment in hardware, the jStart team is working to create lighter-weight configurations of Watson, applicable for different scenarios.
- User Interaction/Experience. Another goal for the team is to enhance how Watson interacts with users. When on Jeopardy!, Watson actually had very limited interactive capabilities--it concentrated solely on answering the Jeopardy questions rather than interacting with the players. The team is working on making Watson more interactive by allowing it to ask questions of users, understand answers (see DeepQA, above), but also understand the context of those questions and answers. More to come later on this exciting new dimension to Watson...
- Machine Learning. Looking at the distant future, plans call for Watson to be a system that not only can take guesses as to what it thinks are the right answers, but can learn from its mistakes. This self learning system adds a degree of sophistication to Watson never before seen when coupled with its other capabilities...the more you use Watson, the "smarter" it becomes.
And that just represents the major components of Watson that are currently being worked on, are on the Watson development roadmap, or that are being asked for by clients which we have already begun conversations with. As you can see, Watson will truly represent a quantum leap in how intelligent systems will be leveraged to build an even Smarter Planet.
IBM Watson Content Analytics is a powerful analytics solution that helps organizations aggregate, analyze and visualize massive amounts of information to expose unique insights. Watson Content Analytics helps organizations interpret and understand their enterprise information to validate what is known and reveal what is unknown. jStart can assist your organization harness the power of Watson to solve current business challenges. If you are interested in talking to jStart about Watson and how it might be used to accelerate your business, feel free to contact us.
Resources and Links
Case Studies and Scenarios
Want more information?
Want to know more about these jStart topics?