Loading lesson…
A computer that played a trivia game show became the face of AI for a moment, then taught a hard lesson about hype.
In February 2011, IBM's Watson played a two-match Jeopardy exhibition against champions Ken Jennings and Brad Rutter. Watson won decisively. Jennings wrote his Final Jeopardy answer with a note: I for one welcome our new computer overlords.
Watson was not a neural network. It was a massive ensemble of classical NLP techniques, hypothesis generation, evidence scoring, and scalable retrieval over encyclopedic text. It ran on an IBM cluster with 2,880 POWER7 cores and 16 terabytes of memory.
IBM immediately pivoted Watson into healthcare, promising AI oncology tools that would transform medicine. The story of Watson Health over the next decade was far less triumphant. Hospitals found the tools hard to integrate, results underwhelming, and IBM eventually sold much of Watson Health in 2022.
I for one welcome our new computer overlords.
— Ken Jennings, Final Jeopardy, 2011
The big idea: winning a game show and transforming medicine turned out to be very different problems. The public saw one and got sold the other.
15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-history-watson-2011-builders
In what year did IBM's Watson compete on the game show Jeopardy?
Which two champion Jeopardy players did Watson compete against?
Which of the following best describes the type of AI system Watson was?
What was unusual about Watson's answer in a US Cities category clue?
How many POWER7 cores did Watson run on during the Jeopardy matches?
What problem did Watson sometimes exhibit during gameplay that revealed a limitation?
After the Jeopardy matches, what industry did IBM immediately try to apply Watson's technology to?
What happened to IBM's Watson Health division by 2022?
What did Ken Jennings write on his Final Jeopardy answer sheet?
What was the main reason Watson struggled with Jeopardy's clues?
What did Watson's odd betting behavior reveal about its decision-making?
What is the name of the question-answering technology that Watson used?
What fundamental lesson does the Watson story illustrate about AI?
What was wrong with Watson's hardware setup in terms of its processing power?
Why did hospitals find Watson Health tools difficult to use?