Cognitive computing enables people to generate value through computing in a new way. Bernard Marr writes “The goal of cognitive computing is to simulate human thought processes in a computerized model.” An alternative to a structured approach to computing – cognitive computing approaches problem solving by mimicking the human approach.
- Observing visible phenomena and bodies of evidences,
- Drawing on knowledge to develop hypotheses,
- Evaluating which hypotheses are correct, and
- Choosing the best option.
Highlights of Watson’s capabilities include:
- interpreting unstructured data,
- interpreting natural language,
- understanding context, and
- learning jargon.
To prepare Watson for solving a problem:
- a corpus of relevant information is loaded into Watson – called a Curation of Content,
- data is pre-processed or Ingested,
- Watson is trained by experts through question-answer pairs, and
- Watson learns through interaction.
Watson relies on deep learning algorithms. The algorithms interpret a training set of examples to learn how something works. Each is example is an input-output pair. A pair might be a picture of a basketball and a picture of a hoop. We show the computer which object is a basketball. Then show it subsequent pictures and ask it to decide if the pictures are of basketballs. The system compares each image with that of a basketball and make an answer.