IBM Watson is a computer system capable of answering questions posed in natural language, developed as part of IBM's DeepQA project under principal investigator David Ferrucci. Named after IBM's founder Thomas J. Watson, the system was designed to compete on the quiz show Jeopardy!, where it faced champions Brad Rutter and Ken Jennings in a televised match in February 2011. Watson won the first-place prize of US$1 million, demonstrating significant advances in artificial intelligence and question answering.
Watson's architecture integrated multiple technologies, including natural language processing, information retrieval, knowledge representation, automated reasoning, and machine learning. IBM stated that Watson used "more than 100 different techniques" to analyze natural language, generate hypotheses, and rank evidence. The system was built on the DeepQA software and the Apache UIMA framework, written in Java, C++, and Prolog, and ran on SUSE Linux Enterprise Server 11 with Apache Hadoop for distributed computing.
Hardware and Performance
Watson employed a cluster of ninety IBM Power 750 servers, each with a 3.5 GHz POWER7 eight-core processor and four threads per core, totaling 2,880 processor threads and 16 terabytes of RAM. The system processed 500 gigabytes per second, equivalent to a million books, and had a Linpack performance of 80 TeraFLOPs. Its hardware cost was estimated at about three million dollars. For the Jeopardy! game, all content was stored in RAM because hard drives would be too slow to compete with human champions.
Data and Knowledge Sources
Watson's knowledge base included encyclopedias, dictionaries, thesauri, newswire articles, and literary works, along with databases and ontologies such as DBpedia, WordNet, and YAGO. The IBM team provided millions of documents to build its knowledge. Watson parsed questions into keywords and sentence fragments, executing hundreds of language analysis algorithms simultaneously. The more algorithms that independently found the same answer, the higher the confidence.
Strategy and Comparison with Humans
Watson included strategy modules for betting and Daily Double detection. One module calculated Final Jeopardy wagers based on confidence scores and contestant standings. Another used Bayes' rule to estimate the probability of unrevealed Daily Doubles, with wagering computed by a two-layered neural network similar to TD-Gammon. Watson's reaction time on the buzzer was faster than humans when reacting, though it could not anticipate the ready signal. It lacked human context understanding but was immune to psychological tactics. In mock games, humans used Watson's six-to-seven-second processing time to buzz first, but in the actual match, Watson's speed and accuracy prevailed.
Legacy and Commercial Applications
In February 2013, IBM announced Watson's first commercial application for lung cancer treatment decisions at Memorial Sloan Kettering Cancer Center, in conjunction with WellPoint (now Elevance Health). However, the Watson Health division was divested in 2022 and sold to Francisco Partners for $1 billion, after costing $4 billion to develop. By 2023, Watson had contributed to IBM losing 10% of its stock value and mass layoffs. Despite these setbacks, Watson's 2011 Jeopardy! victory remains a landmark in Artificial intelligence and Machine learning, influencing subsequent developments in Deep learning and Natural language processing.