IBM Watson partnering people with #CognitiveComputing
Posted on November 17th, 2014
11/17/2014 @WeWork, 69 Charlton Street, NY
David Melville, @IBM research spoke about #Watson and its potential uses as a collaborative tool. Watson is famous as a contestant on #Jeopardy. The basic processes it undertakes can be classified:
- multiple parsings of a questions
- generate hypothesis
- collect & evaluate evidence – e.g. frequency of a word in a doc, source reliability, context of work within a document
- weight & combine for final confidences
The engine is written in Java and basic system is open source, but parts of the annotator are proprietary. More technical information is available for developers at http://bluemix.net.
David showed videos demonstrating 1. Fast responses (Jeopardy) 2. Data retrieval (document summary followed by search) 3. Simple learning 4. Speech recognition
The talk was not technical with the emphasis on how the capabilities demonstrated via the videos could be used to create human-computer collaboration.
It is unclear what Watson contributes. It is fast and it can mine large databases. However, much of the expertise which it can contribute would be in specialized areas where human experts contribute both “book knowledge” in depth, but also creative insights. Watson’s current creativity is probably limited and human “experts” are primarily experts because they know the topic area from experience in detail beyond what is available in documents. Watson could be “an expert in a box”, but will probably require time before it can replace a human expert who not only knows the facts, but can integrate the facts with the human/business aspirations and limitations. In these cases, speed helps, but is not the deciding factor in making good decisions.
Speed might be the main contribution of Watson. For instance, Watson’s data retrieval could be used in high speed trading or other activities which require super-human reaction times. Other uses could be in autonomous drone attacks, extra terrestrial exploration, or other activities which require a breadth of knowledge whose success is highly dependent on speed and autonomy. In these cases applying rules quickly is key and Watson can handle a large number of complicated rules.
The use of Watson may hinge on whether the task requires knowledge or wisdom and the speed at which we need to decide.