IBM Watson – The Science behind an Answer and the Era of Cognitive Computing

 

2011-04-ibm-watson-1

Watson – 2880 CPU cores, 90 Power7+ nodes, 16 TB aggregated main memory, knowledgebase of two hundred million book pages and a sophisticated semantic query engine.

I assume you all tracked the Jeopardy challenge which Watson won on the year 2011. The accuracy with which Watson was able to answer the questions was amazing and we should credit the members of IBM Research – Watson Lab for inventing DeepQA, a system able to understand the semantics of human language.

This is a huge difference to traditional search engines which only can interpret text queries and don’t understand the actual semantic meaning.

So, for example the query: “Which disease causes ‘Uveitis’ in a patient with family history of arthritis presenting circular rash, fever, and headache?”, a traditional search engine would answer with a set of links to web pages which a domain expert then has to read through in order to get the relevant information.

If you ask the same question to Watson, the answer would be:

76% Lyme disease,

1% Behcet’s disease,

1% Sarcoidosis

Besides returning the correct answer, Watson is also able to demonstrate the confidence about the correctness of his answer. This is a new paradigm in Cognitive Computing where machines exhibit non-deterministic behavior – just as human beings. To address the demand of this new type of application IBM recently announced the Watson Group.

But IBM doesn’t want to win game shows – we want to build a Smarter Planet. For example, in 15 seconds, Watson for Healthcare can read two hundred million pages of clinical data, cross-reference the symptoms of one million cancer patients or read millions of current medical journals to test hypotheses.

In order to let humanity benefit even more from Watson, IBM provides Watson in the Cloud in the Watson for Ecosystem program, where, Independent Software Vendors (ISVs) will be able to build their customized solutions based on Watson technology.

Check out this video of the IBM Watson Group launch event on January 9, 2014 in New York.

One prominent example is “The North Face” where Watson improves consumer experience with a solution built by IBM ISV partner FluidRetail.

But, the ability to return results in milliseconds was mainly attributed to the new and innovative way IBM is designing microprocessors. Since a couple of years, CPU speed is growing faster than access speed to main memory. Although multiple layers of cache dampen this problem, a huge bottleneck is emerging. In current and future generations of systems, if this gridlock is not addressed, the CPU is primarily waiting to fetch data from the main memory.

Starting with the Power7 family, IBM introduced DRAM (main memory technology) attached directly to the CPU on the same wafer. IBM calls this technology eDRAM (embedded DRAM).  With the fastest and densest eDRAM in the industry, Watson’s complete knowledge base was just a couple of millimeters away from the Power7+ CPU cores and this finally led to response times in milliseconds.

The fun has just begun! In 2013, IBM founded the OpenPOWER consortium together with Google, NVIDIA, Mellanox and Tyan to bring the Power platform to the broad data center market.

If this appears far too complicated for you, IBM’s Expert Integrated Systems based on Power Technology, either IBM PureFlex System for private IaaS clouds or IBM PureApplication Systems for your private PaaS clouds could hide away the whole complexity of networking, storage, hardware and software in an easy to use interface leveraging the unite power of all components together.


Forrester Consulting White PaperIn a commissioned study, Forrester Consulting describes how the adoption of integrated systems, prompts IT leaders to converge technical roles and adopt new processes for faster solution delivery. Learn how integrated systems can help you deliver solutions faster in this Forrester Consulting Study.

Leave a comment
Romeo Kienzler

About Romeo Kienzler

Romeo Kienzler works as Data Scientist and Architect at the IBM Innovation Center specialized in (Big) Data Management, Applied Statistics and Artificial Intelligence. He holds a Diploma in Distributed Systems and a M. Sc. in Information Systems with specialization in Applied Statistics and Bioinformatics. Before joining the Innovation Center he worked as Software Engineer and Architect at IBM Software Group Switzerland.
This entry was posted in application system, Cloud, expert integrated systems, Hardware, Mobile, Patterns, Performance, Platform Selection, Security, Storage, Support, Virtualization and tagged , , . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>