An inside look at the Power behind IBM Watson

Editor's note: This is a guest post from Tilak Agerwala, Vice President, Systems, IBM Research.

The IBM POWER7 processor behind Watson is specifically designed to meet the demands of workloads, like IBM’s DeepQA natural language processing technology. POWER7 is an ideal system for running thousands of analytical tasks at once, which is what the Watson Deep QA software requires. In order to answer a Jeopardy question in under 3 seconds, the system run thousands of breakthrough analytical tasks at once.

Watson uses POWER7 to deliver massive parallelism of multiple complex tasks that execute simultaneously on individual processor threads. Watson relies on multiple IBM Power 750 servers clustered together, each with four processor sockets with eight POWER7 cores per socket, and four threads per core. Combined, they make a workload optimized system that can answer questions posed in natural language in seconds. No other system in the world can do this.

While the Power 750 server provides excellent capabilities to support the IBM Watson Deep QA software, the Power 750 server was not specially designed for Watson. In fact, the same Power 750 servers are ideal for running many types of analytical tasks and available today to help answer practical business challenges across many industries, such as healthcare, financial services, and call center environments.

Power Systems Driving Progress for a Smarter Planet

Rice University in Houston uses IBM Power 755 systems for their cancer research. The challenge for Rice was to accelerate the understanding of the molecular basis of cancer through the application of genome analysis technologies. The researchers used POWER7 systems to provide more flexibility and efficiency in how they analyzed a range of problems.

“The POWER7 architecture has enabled us to pursue a broader range of research problems on a single system than was possible before.” – Dr. B. Kim Andrews, manager of Research Computing.

At GHY International, a customs brokerage firm in Canada, they needed to be better prepared for clients increasing engagement in international trading. GHY needed to quickly deploy new services – and the infrastructure to support those services. So they migrated to a new Power 750 running Power AIX, Power i and Power Linux. And with PowerVM virtualization, they were able to deploy new capabilities in as little as five minutes to support their new-found growth.

Nigel Fortlage, vice president of IT, said that with PowerVM they went from spending 95 percent of their time managing or reacting to their environment, to now only 5 percent.

Today, Power Systems based on POWER7 processors are already helping thousands of businesses deliver services faster, with higher quality, and superior economics. By harnessing the processing prowess of POWER7, Watson and IBM DeepQA give businesses a new dimension in how to solve their most-complex challenges.


IBM computer system a scientific achievement in natural language processing

Editor's note: This is a guest post from Charles Lickel, Vice President, Software, IBM Research.

The IBM Watson computer system, named after company founder Thomas J. Watson, Sr. catapults data analysis by understanding the nuances of "natural language" that humans use every day.

A breakthrough in the scientific field of DeepQA – a field aimed at furthering computing intelligence – when we began this project more than four years ago, others in the scientific community believed this task to be impossible and we were uncertain about what we would be able to achieve. In fact, it took our research team two years to get the system to be able to analyze a Jeopardy! clue and provide an accurate response in less than three seconds. That was the moment we knew we had a chance of achieving a breakthrough in artificial intelligence and natural language processing, as well as a chance to compete against Jeopardy! champions.

Watson uses Unstructured Information Management Architecture (UIMA), IBM’s open-source framework for analysis of unstructured content, to understand natural language text, speech, images and video. The complex algorithms for the analytics engine behind Watson are tuned specifically for “open-domain” QA – covering a wide domain of knowledge including history, geography, arts, science, sports, and popular culture. It helps Watson handle the broad range of information that language can express and evaluate the “evidence” it collects to determine a confidence level to answer a question.

This fall, the Watson system and the Research team achieved remarkable results, when Watson played more than 55 “sparring games” against former Jeopardy Tournament of Champions contestants.

Watson's ability to understand the meaning and context of human language, and rapidly process information to find precise answers to complex questions, holds enormous potential to transform how computers help people accomplish tasks in business and their personal lives. Watson will enable people to rapidly find specific answers to complex questions. The technology could be applied in areas such as healthcare, for accurately diagnosing patients, to improve online self-service help desks, to provide tourists and citizens with specific information regarding cities, prompt customer support via phone, and much more.

Like Deep Blue, the IBM supercomputer that defeated the reigning world chess champion in 1997, Watson represents a major leap in the capacity of information technology systems to identify patterns, gain critical insight and enhance decision-making despite daunting complexity. But while Deep Blue was an amazing achievement in the application of compute power to a computationally well-defined and well-bounded game, Watson faces a challenge that is open-ended and defies the well-bounded mathematical formulation of a game like chess. Watson has to operate in the near limitless, ambiguous and highly contextual domain of human language and knowledge.

Win or lose, Watson will open eyes to what kinds of questions a computer can answer, and open doors to what kind of problems a computer can solve. For more about the technology behind Watson, visit ibmwatson.com.


Congratulations to IBM's newly appointed ACM Fellows

Each year, around 40 world leaders in science and technology are part of an elite group that are selected to be named ACM Fellows. The distinction of ACM Fellow comes to those achieving accomplishments that are driving the innovations necessary to sustain competitiveness in the digital age. ACM Fellows are born out of the world's leading universities, corporations and research labs and serve as distinguished colleagues to whom the ACM and its members look for guidance and leadership as the world of information technology evolves.

IBM has held steady in ACM Fellow appointments and this year is no exception. Of the 41 ACM Fellows named for 2010, two IBM researchers achieved this distinction: David Ungar and Shumin Zhai.

David Ungar, selected for his contributions to the design and implementation of object-oriented programming languages, reports in to the Programming Models and Tools Research group at IBM Research - T.J. Watson in NY. David is credited for work on "Tuning Fork," a performance analysis and visualization tool and "Renaissance" - in pursuit of a breakthrough in productivity and performance for massively parallel, tightly-coupled systems such as manycore CPUs. Adding to his list of achievements, David led a group that invented technology that made Java practical and co-designed a programming language that demonstrated how to simplify object-orientation while increasing expressive power. Think JavaScript; it was at the very least the inspiration for the person who designed it.

Having come a long way from repairing stereos and TVs and maintaining and analog computer, David is an expert recognized by peers far beyond IBM.

Shumin Zhai, an IBM Research - Almaden dynamo, was honored for his contributions to human-computer interface research and innovation. Side note: Zhai was a university professor at age 23.

Zhai explains, "I grew up during China's culture revolution, so no one had a job in today's sense. In 1977 - at age 16 - I was among the one percent lucky national university entrance examinees, the first such group after the 10 year culture revolution, to enter universities." He goes on, "By the time I finished my master's degree and started teaching, I was still not much older than the university students I taught. So when I walked into my first lecture I was visibly startled by the students who all customarily raised then burst into laughter. But by the end of the semester, they elected me "the best lecturer" among a teaching staff of well over 100."

Now, as an IBMer, Zhai has several accolades under his belt. We have Zhai to thank for helping invent the ScrollPoint mouse and other input devices of IBM PCs and Thinkpads. He pioneered the touch screen gesture keyboard input method (smartphone, anybody?) with the ShapeWriter project. He's now advancing that research by discovering and refining models and empirical laws of user action on computer screens which quantitatively predict users' performance in pointing, crossing, steering, and stroke gesturing.

Ungar and Zhai join a list of several accomplished IBMers named ACM Fellow over the last 15+ years.


IBM scientist becomes a Humboldtian

Dr. David DiVincenzo, a quantum physicist from the IBM Thomas J. Watson Research Center in New York, has been awarded the Alexander von Humboldt International Award for Research in Germany. Named for the German naturalist and explorer, winners of the award, also know as Humboldtian's, will have access to a grant worth up to five million Euros to conduct cutting-edge research at German universities.

Dr. DiVincenzo, who researches  quantum information theory and quantum computing, was nominated by RWTH Aachen University and Forschungszentrum Jülich, the later being home to the 9th most powerful supercomputer in the world, an IBM Blue Gene.  

Dr. DiVincenzo said, " I am pleased that Aachen and Juelich have recognized the future of information technology as one of the centerpieces of their research agenda.  I expect to have a productive contribution in this area, as well as in the broader area of theoretical nanoelectronics." 

The other winners include:
  • Psycholinguist Harald Clahsen, currently researching at the University of Essex, Colchester, UK, is to work at the University of Potsdam.
  • Vahid Sandoghdar is a physicist at the Swiss Federal Institute of Technology Zurich and is to conduct research for the University of Erlangen-Nürnberg and the Max Planck Institute for the Science of Light in Erlangen.

Commenting in the latest winners, Helmut Schwarz, the Foundation's President said,  “The nominations from the successful universities of Aachen, Erlangen-Nürnberg and Potsdam show how the Alexander von Humboldt Professorship may be used as a strategic instrument: The award winners will help to expand existing areas of excellence and intensify co-operation with non-university research institutions. The award winners’ knowledge and international repute will contribute to the fulfilment of these strategies.”

Every year, the Alexander von Humboldt Foundation enables more than 2,000 researchers from all over the world to spend time researching in Germany. The Foundation maintains a network of well over 24,000 Humboldtians.

The award ceremony is scheduled for 12 May 2011 in Berlin.