To Think: Watson and the Open Question

Deep Blue could only play chess. The next milestone had to handle something messier: an open-ended human question, asked in ordinary language.

Key Milestones

• IBM Watson defeats the Jeopardy! champions, 2011. Watson, a natural-language question-answering system, defeated the show's two greatest all-time champions, Ken Jennings and Brad Rutter, with no connection to the internet during play.

The Energy, Social, and Economic Impact of Thinking

The final scores were not close. Watson finished with $77,147, more than three times either Jennings's $24,000 or Rutter's $21,600 on their own. What made Watson different from Deep Blue was not the margin, it was the domain. Chess has fixed rules and a finite board. Jeopardy! clues involve puns and ordinary human phrasing, exactly what brute-force search cannot calculate its way through.

View chart in the full report

Watson's win, and the $1 million prize IBM donated to charity, marked the moment a machine first parsed an ordinary question and retrieved a correct answer from a vast, unstructured body of knowledge, not just a rulebook. That is a meaningfully different arbitrage: not brute calculation, but broad, general recall.

Little did we know that IBM had led us into a new era of energy arbitrage of the mind, by way of a machine that could play chess and Jeopardy!, compute and understand language, and recall what it had learned.

We now give this new form of energy arbitrage a shorter name: AI.