Over the past few years, as the S&P 500 jumped nearly 60 percent, IBM’s stock sunk 6 percent. Big Blue needed a miracle — and quickly – if it wanted to compete for top tech talent. Enter Watson, the computer that took down “Jeopardy” champion Ken Jennings, and IBM’s best hope for the future.
Watson began as a research project for a few engineers, functioning like a startup within IBM’s behemoth walls. Today, IBM Watson employs thousands of team members who work not only to program but also monetize Watson. In a company panned by financial analysts as an obsolete, poorly run dinosaur, the internal startup has become a source of new life.
IBM used to be so committed to maintaining a specific earnings-per-share ratio that it doesn’t reinvest in innovation. Instead, IBM lays off employees, partners with cheaper technical talent, and buys companies instead of innovating from within.
A decade ago, IBM was no stranger to competition. Its Deep Blue chess-playing computer defeated chess icon Garry Kasparov in 1997. Unfortunately, no one was in the market for a computer that did nothing but play chess. IBM needed a product with both sex appeal and a hungry target market.
One night, a research manager at IBM, Charles Lickel, found himself eating dinner in a silent restaurant. His fellow patrons had their eyes glued to a television set, watching “Jeopardy” über-champ Jennings cruising along on his 74-game winning streak. Lickel had the insight that if IBM could make a computer that could defeat Jennings, it would be a major publicity win for the company. By 2005, he’d engaged David Ferrucci, IBM’s senior manager for Semantic Analysis and Integration, to launch the Watson project.
Computers that could process natural language existed in 2005. They could parse documents and answer simple questions about the content, but they couldn’t pick up on the intended meaning behind language.
For a New York Times profile, Ferrucci gave the example of a “Jeopardy” clue, like “The name of this hat is elementary, my dear contestant.” The human mind would understand the clue was related to Sherlock Holmes, but a computer thinking mathematically and literally wouldn’t understand the pun. Even a question like, “What kind of hat does Sherlock Holmes wear?” couldn’t be answered unless the computer could find information stored as a precise answer to that specific question. Added to these challenges was the speed required to spit out a “Jeopardy” answer. Jennings’ artificial challenger would have to come up with a correct answer a split second after the buzzer.
The problem demanded a scientist who wasn’t afraid to take on a big challenge. As Ferrucci said, “I had no interest spending the next five years of my life pursuing things in the small.”
By 2007, Ferrucci, who had a team of 15 researchers, saw a huge advantage in IBM’s supercomputing power. incredible server capacity made it possible to feed incredible amounts of information to Watson. It also gave Watson the computing power to simultaneously analyze hundreds of complex statistical-linguistic algorithms.
While competitors were trying to create one perfect algorithm for analyzing language, Ferrucci realized Watson needed to think like a human. It would need hundreds of approaches to solving language problems, not one perfect language processing formula. As a result, Ferrucci and his team programmed Watson with over 100 algorithms for processing natural language. They also programmed Watson to use probability to guess at correct answers, meaning it chose the best answer instead of taking every word literally. By 2011, Watson became master of “Jeopardy.” In 2014, Watson had spawned its own business unit.
Creative thinkers need room to waste money and make a few mistakes, functions that are anathema within most big corporations. IBM’s old, poor decision-making and the change to allow the creation of Watson would make a great case study for anyone who wants to complete a graduate finance degree. By giving Ferrucci latitude to experiment with Watson, IBM created a powerful potential revenue source. Its applications are shaping the future of health care, finance, and other sectors.
Within a decade, IBM CEO Virginia Rometty predicts Watson will generate $10 billion in revenue for IBM. It’s success serves as a lesson for any enterprise growing stagnant in today’s market. Instead of buying someone else’s startup, try launching your own.