Elementary, My Dear
Answer: Second-year Texas Evening MBA student Toan Pham helped build the technology for this revolutionary machine that recently wowed the world with its artificial intelligence.
If you guessed, “What is Watson?” you are correct! No prize money involved here, but at least you learned something new.
TEMBA student and IBM engineer Toan Pham worked on the main processing unit used to power Watson, the computer that won a three-day game of Jeopardy! against the game’s two most successful human players.
Trent Thurman: Tell me about your specific involvement with the Watson project.
Toan Pham: I worked on the P7 chip. This chip is the main processing unit of the system that is Watson. My work involved creating circuit for the L2 cache unit and making sure the unit functioned at the desired frequency. The L2 cache unit acts as a fast, temporary storage area for the processor. You can imagine the L2 cache as a filing cabinet, versus main storage being a library. It is much quicker to look up information in a filing cabinet than searching through a library.
TT: When did you first learn that the CPU you are working on would be a part of the Watson project?
TP: We first heard about this a month ago in an internal memo. When we started the project five years ago, we knew one application of this CPU would be in high-end computing systems for the government. Typically, our product goes into a six-foot tall metal box that goes into an air-conditioned room, never to be seen again until its replacement time comes. There is not a lot of glitz and glamour. The Watson project demonstrates that serious high-end computation can be glamorous and fun when used in the right application.
TT: Question answering (QA) systems obviously have very practical business applications and IBM has already announced some of the plans. Where do you personally see this technology going?
TP: I see this technology having the potential to change how we interact with computers. An important advancement is how the computer is able to interpret language and select the key words. This intelligence can allow us to interact with the computer through speech instead of a keyboard and mouse. We can dictate instead of using a word processing program. We can specify what we want instead of creating slide decks. The technology can be enhanced further to aid with interactive computing. Imagine proposing a hypothesis and the computer would cross reference past occurrences, perform the analysis, and present the solution. One specific example is business strategy exploration.
TT: Outstanding. Tell me about the mood in the office over the last few days.
TP: The mood has been very upbeat. IBM held a three-day Jeopardy! watching marathon, and all employees were encouraged to attend the celebration and check out Watson’s performance. The hallway was buzzing with speculation of Watson’s algorithm, implementation, future applications, as well as Watson’s quirks. This event has been a great morale booster for everyone.
TT: UT, along with seven other universities, researched and developed a lot of the intelligence for Watson. Did you have any involvement with these groups?
TP: Regrettably, I have not had any involvement with these groups.
TT: What’s up with the random $6,435 wager on the daily double?
TP: I believe Watson’s betting algorithm is quite complex. As in any financial decision, it has to examine the risk-reward trade-off while taking into account the question’s category, its past performance and the competition’s position. Humans are comfortable with ranges and variation so we tend to round off. Watson, like any computer, is comfortable with being exact.
TT: Makes sense. OK, one more. Toronto? (Under the category “U.S. Cities,” Watson responded with “What is Toronto?????” when given the clue, “this city’s largest airport is named for a famous World War II hero, its second largest for a famous World War II battle.”)
TP: Well, there are conspiracy theories…I’ll share one of my own. I think Watson purposely gave the wrong answer for the first day’s final question. Knowing that it was so far ahead of the competition, Watson wagered practically nothing and got the answer wrong so the competition could “catch up”. Compare that to the second day’s final question where Watson was neck and neck with Ken Jennings. Watson wagered most of its earnings, then correctly answered an arguably more difficult question.