Andre Standback IT 103, Sec. 001
2/21/12 “IBM’s Watson”
"By placing this statement on my webpage, I certify that I have read and understand the GMU Honor Code on http://academicintegrity.gmu.edu/honorcode/ . I am fully aware of the following sections of the Honor Code: Extent of the Honor Code, Responsibility of the Student and Penalty. In addition, I have received permission from the copyright holder for any copyrighted material that is displayed on my site. This includes quoting extensive amounts of text, any material copied directly from a web page and graphics/pictures that are copyrighted. This project or subject material has not been used in another class by me or any other student. Finally, I certify that this site is not for commercial purposes, which is a violation of the George Mason Responsible Use of Computing (RUC) Policy posted on http://universitypolicy.gmu.edu/1301gen.html web site."
Introduction
“Watson” is an artificial intelligence supercomputer that is designed to be a “question-answering” machine. This supercomputer was created by International Business Machines Corporation (IBM Corp), a multi-national technology and consulting corporation. The most recent challenge undertook by IBM was to develop a computer system that could compete with other human contestants on the popular game show called “Jeopardy.” This idea started in 2004 for IBM after the interest with “Deep Blue,” which is a supercomputer that competed with expert chess players. IBM’s vice president was interested in the game show Jeopardy, and in particular, the shows champion contestant Ken Jennings and his large winning streak. The Jeopardy
Challenge, or the competition between Watson and humans, became formalized in 2006 as IBM researchers conducted a feasibility study to examine this opportunity. The project officially began in 2007. Jeopardy is a quiz show consisting of an answer-and-question format of trivia questions in the areas of science, geography, literature and various other subjects. This challenge subsequently led to the development of the DeepQA project, which was initiated for the purpose of advancing computer technology to rival the previous Deep Blue project.
Architecture
The architecture behind Watson is very complex which is necessary in order to for the computer to understand natural human language in the same way that two human beings
communicate with each other. The field of question-answering computer systems has advanced over the years, and some may wonder what makes Watson different from any other search engine. A search engine requires precision when asking a question, but Watson would have to operate successfully without the normal precision typed searches and interpret ‘spoken’ language questions on Jeopardy. Also, Watson would not be connected to the Internet, so only a limited
amount of memory could be used to retain all of the data that Watson needed for the purpose of analyzing the question and coming-up with an answer. The basis for Watson to understand language lies in natural language processing, semantic analysis, information reasoning and machine learning. Watson had to be able to understand a question, generate ideas, make a hypothesis and check that hypothesis from credible resources, verify confidence of the
hypothesized answer generated and submit that answer -- all in a few seconds or less. In order to understand human language, Watson runs on a Linux operating system -- the same operating system that runs almost all supercomputers. According to Bill Pulleyblank, director of
Exploratory Server Systems at IBM Research, with regard to the reason for selecting Linux was “Because we believed it could be extended to run a computer the size of Blue Gene. We saw considerable advantage in using an operating system supported by the open-source community, so that we can get their input and feedback” (Ping, 2009). The Linux operating systems allows Watson to operate at incredibly fast speeds due to the ten racks of IBM POWER 750 servers, 15 terabytes of RAM, 2,880 3.5GHz POWER7 processor power cores at a speed of 80 Teraflops. To give meaning to the speed of 80 Teraflops, an iPad 2 with an A5 Processor runs at an
estimated 0.000171 Teraflops. Watson also has various software components that allow it to run at peak efficiency. Most of the software is available for purchase at any retail store that sells computer equipment and software such as Best Buy. Watson is programmed in Java and C++, and uses Apache Hadoop and Apache UIMA frameworks. The Apache Hadoop allows Watson to use a Map Reduce algorithm that breaks down very large tasks to smaller and simpler
components. This algorithm is especially helpful when searching is personal database for answers. Apache UIMA is similar in the way of making large tasks smaller but also allows Watson to make of use of data presented in English text, audio or video files. Apache UIMA
was the driving force in developing an answer. Application software alone was not a sufficient enabler to help Watson with this tremendous task of playing in Jeopardy, so Watson is also equipped with structured and non-structured databases from the complete text of Wikipedia, and multiple encyclopedias, dictionaries, thesauri, articles, literary works and access to other data from DBPedia, WordNet, and YAGO. WordNet acts as a “super thesaurus” for Watson, and DBPedia functions as a more structured form of Wikipedia.
Jeopardy Challenge
In 2010, before Watson could play against the champions of Jeopardy, it needed to demonstrate its worthiness by playing against other experienced contestants. These practice games were located at IBM Research in Yorktown Heights, New York. As a passing criterion, Watson also needed to win about 70 percent of the games against former Jeopardy champions. February 2011 was when Watson finally made its debut on national television competing against Jeopardy champions Ken Jennings who had won 74 consecutive matches at that time and Brad Rutter who was the biggest money winner at $3.2 million and had never been defeated. On Day-1 of the competition, Watson got an early lead and was doing well. The other two contestants eventually closed this gap by the end of Day 1 with Watson and Rutter tied at $5,000 and Jennings at $2,000. On Day-2 of the competition, Watson achieved a massive lead on the other two contestants, most notably brought about when Watson got two daily-double questions
correct. Watson was on a roll, and Rutters and Jennings were unable to break its winning streak. The contestants had to make a wager for the final jeopardy question of this round and for the first time, Watson got the question wrong, while both Jennings and Rutters got it right. Watson, however, displayed a lack of confidence before answering the question and was smart enough to only wager $947, which strategically, didn’t change its position as the leader. At the end of this
round Jennings was at $4,800, Rutters was at $10,400, and Watson had a substantial lead of $35,734. This was also the round wherein Jennings started to show frustration as Watson was much faster at the buzzer, and consequently, all the questions were directed to Watson. On Day-3, the contestants each started with a score of zero. Watson still dominated the competition, despite trailing behind Jennings for a short while. The last jeopardy question for the entire challenge had two memorable moments: (1) Jennings jokingly stated under his written answer “I for one welcome our new computer overlords,” (xMysticKnightx, 2011) and (2) Watson wagered a total of $17,973 for the last question which brought about Watson’s huge landslide victory after all the points were totaled up from all three rounds. Watson won the Jeopardy challenge with $77,147 while Jennings was in second with $24,000 and Rutters was third with $21,600.
The Future
IBM believes that the technology behind Watson can now be used for a variety of other applications, specifically in the field of medicine. The plans to get Watson and other future supercomputers like it into the field of medicine has already began as evident by the new partnership between WellPoint Inc. and IBM. WellPoint Inc. is the nations largest managed health care company of the Blue Cross and Blue Shield Association. Both companies agree that having Watson in the medical field would be an advantage for physicians because the
supercomputer will be able to sift through millions of pages of data and suggest a diagnosis for each patient case almost instantly. Lore Beer, WellPoint’s Vice President stated “We see this as a tool to help them be more successful in driving better outcomes for our members. We’re not trying to replace the physician” (Helfand, 2011). WellPoint has also attempted to assure patients by stating that the computer won’t but used to supplant doctors, but rather is a tool to provide a quality answers for medical diagnoses. The two problem areas in health care where Watson will
be most beneficial are the reduction in quantity of medical diagnosing errors and alleviating the strain caused by the shortage of physicians available. In a report by the Institute of Medicine, there are about 98,000 deaths every year because of medical errors, which are ranked as the fifth leading cause of death in the U.S. Watson can hold huge amounts of data from various medical texts and will be able to rapidly analyze the data without shortcuts or biases. The Association of American Medical Colleges predicts that there will be a shortage of 150,000 physicians in the next 15 years. Watson would alleviate this shortage by allowing non-physician health care providers assume a greater role.
Conclusion
The contest of Watson vs. ‘human contestants’ was never about winning in Jeopardy, nor was it about creating a self-aware computer like HAL 9000 from 2001: A Space Odyssey. If anything, the competition was a test to push Watson’s limits and capabilities to explore its potential in other applications. The goal of this whole project was to create something similar to the ship’s computer in Star Trek: The Next Generation. The project was a huge success for IBM, and it was a quantum leap for a machines’ ability to understand context within the human language.
References
Ferrucci, D., Brown, E., Chu-Carroll, J., … Welty, C. (Fall, 2010). Building Watson: An
Overview of the Deep QA Project. Retrieved February 23, 2012, from the Association for the Advancement of Artificial Intelligence website:
http://www.aaai.org/Magazine/Watson/watson.php
Is a complete journal article from an online magazine that describes the artificial
intelligence behind Watson and how the software and hardware works against the game Jeopardy.
Heaton, J. (2011, March 12). The Free and Open Software Behind IBM’s Jeopardy Champion Watson. Retrieved February 23, 2012, from 2005-2012 Heaton Research, Inc. website: http://www.heatonresearch.com/content/free-and-open-software-behind-ibm’s-jeopardy-champion-watson
Heaton Research is a technical book publishing company that specializes in computer hardware, networks, and other computer related applications. This article was an analysis behind the software that operates the super computer Watson.
Helfand, D., (2011, September 12). IBM’s Watson supercomputer to give instant medical diagnoses. Retrieved February 24, 2012, from LA Times Business website: http://latimesblogs.latimes.com/technology/2011/09/ibm-watson-wellpoint.html
This is an online news article from the LA Times that discusses other applications for the supercomputer Watson, primarily in the field of medicine and other future prospects. Kanellos, M. (2002, October 28). IBM’s “Blue Gene” to run on Linux. Retrieved February 23,
This is an article from ZDNet, which is a business technology news website that is published by CBS interactive. Despite the article being about “Blue Gene,” a different supercomputer from IBM, it discusses why all supercomputer by IBM uses a Linux operating system.
Paul, I. (2011, February 17). IBM Watson Wins Jeopardy, Humans Rally Back. Retrieved February 24, from 1998-2012, PCWorld Communications, Inc. website:
http://www.pcworld.com/article/219900/ibm_watson_wins_jeopardy_humans_rally_back .html
This article comes from the PC World magazine, which is a global computer magazine that is published by IDG that offers advice about computers, the internet, and other related topics. The article discusses the results of Jeopardy between the supercomputer Watson and the other human contestants
Pepitone, J. (2011, January 14). IBM’s Jeopardy supercomputer beats humans in practice bout. Retrieved February 23, 2012, from 2012 Cable News Network. A Time Warner Co. CNN Money website:
http://money.cnn.com/2011/01/13/technology/ibm_jeopardy_watson/index.htm This article comes from CNN in the money section, a U.S. cable news network. This article describes how the supercomputer Watson performed during the Jeopardy game show and possible future applications for Watson
Pingom, (2009, June 24). The triumph of Linux as a supercomputer OS. Retrieved February 23, 2012, from 2011 Pingdom AB website:
Pingdom is a website monitoring service. This article comes from a blog on Pingdom’s website that discusses why the operating system Linux is used for supercomputers. It also displayed a chart of the top 20 supercomputers in the world and that 19 out of the 20 used Linux.
Suh, Y. (2011, February 8). ‘Watson’ could transform medicine. Retrieved February 24, 2012, from 2012 USA TODAY Gannett Co. Inc. website Forum:
http://www.usatoday.com/news/opinion/forum/2011-02-09-column09_ST_N.htm
USA Today is a national American daily newspaper published by the Gannett Company. This article comes from the opinion section which discusses the future possibilities of using the supercomputer Watson in the field of medicine, and the implications behind that act.
xMysticKnightx (2011, February 16). Jeopardy! IBM Watson Day 1 (Feb 14, 2011) Part ½ … Jeopardy! IBM Watson Day 3 (Feb 16, 2011) Part 2/2 [Video Files Day 1-3]. Retrieved February 24, 2012, from: http://www.youtube.com/watch?v=seNkjYyG3gI&context=C39e9620ADOEgsToPDskI KqUWKrX3Yt_62OllMufAV http://www.youtube.com/watch?v=39AtGQgJgsI&context=C39e9620ADOEgsToPDskI KqUWKrX3Yt_62OllMufAV http://www.youtube.com/watch?v=4nt-abT8gKw&context=C39e9620ADOEgsToPDskIKqUWKrX3Yt_62OllMufAV http://www.youtube.com/watch?v=-QYchgv5dMM&context=C39e9620ADOEgsToPDskIKqUWKrX3Yt_62OllMufAV
http://www.youtube.com/watch?v=iGDI-SD9Cbo&context=C39e9620ADOEgsToPDskIKqUWKrX3Yt_62OllMufAV
http://www.youtube.com/watch?v=o6oS64Bpx0g&context=C39e9620ADOEgsToPDskI KqUWKrX3Yt_62OllMufAV
These are a list of video clips retrieved from the online video-sharing website YouTube. Together these clips are a recording of the game show Jeopardy that the supercomputer Watson performed in.
Vaughan-Nichols, S.J. (2011, January 13). I, for one, welcome our Linux Penguin, Jeopardy Overlords. Retrieved February 23, 2012, from 2012 CBS Interactive ZDNet website: http://www.zdnet.com/blog/open-source/i-for-one-welcome-our-linux-penguin-jeopardy-overlords/8081?tag=content;siu-container
This article like the others comes from ZDNet and continually discusses why Linux is the superior operating system for supercomputers, primarily in regard to the supercomputer Watson