To replicate Watson's Jeopardy performance, what capabilities would a program need?

• A. Natural language processing, large knowledge retrieval, and rapid reasoning
• B. Simple arithmetic and manual data entry
• C. Visual pattern recognition only
• D. Mechanical proprioception only

Answer: (A)

To replicate Watson’s Jeopardy performance, a program must combine natural language understanding, broad knowledge access, and fast, evidence-based reasoning. Clues on Jeopardy are posed in natural, often playful language, so the system needs to parse the sentence structure, grasp nuances, and map wordplay to potential facts. Access to a huge knowledge base—across history, science, literature, pop culture, and more—and efficient retrieval means the program can pull relevant information quickly to consider candidate answers.

Rapid reasoning ties it together: the system must generate multiple plausible hypotheses from the clue, weigh the supporting and opposing evidence from the retrieved material, and score each possibility under time pressure to choose the best answer. This involves not just knowing facts, but evaluating which fit the clue most convincingly and aligning with the required Jeopardy response form.

Nan-arithmetic tasks or manual data entry can't address the linguistic understanding or the breadth of knowledge needed, and visual pattern recognition or proprioception wouldn’t help with answering text-based clues at speed.