David McQueeney, VP of Software at IBM Research, dropped in on the theCube for a discussion on Watson and other projects.  McQueeney covered the range of IBM’s worldwide software research in the discussion, sharing history, goals, and accomplishments along the way.  The video can be seen here.

Watson and Deep Blue

On the topic of what business reasons had brought Watson to light, McQueeney talked about going back to the infamous Deep Blue chess match, where a machine beat a famous chess master.  The use case being that IBM at the time felt computing had advanced to the point that analytics could have a lot more applications than they were seeing, and they were looking for a way to demonstrate reasoning and strategy.  This accomplishment fired a lot of people’s imaginations.  This led to the next grand challenge we saw in the Jeopardy challenge and win.  In this realm there were nuances that had to be overcome, in language and context.  This was a formidable task in that rules rarely exist and do change over time.  One of Watson’s key accomplishments was the ability to utilize over 200 million pages of content for each question and respond in seconds.

Further applications

Undoubtedly a unique and seminal moment in time, the conversation turned to its further applications, social impact, and additional uses.  McQueeney discusses one such application where in the medical industry, there is a potential of savings in the billions of dollars.  Crossing that bridge from research into real life, it is an advantage that Watson wasn’t created just to get to Jeopardy; it was always aimed at practical applications.  In fact, Jeopardy was just an initial accomplishment, in many ways easier than other scenarios.  Looking at medical applications, the goal is not to replace doctors with a machine.  The system adds value in that it can help doctors in new ways that couldn’t be done before.  There’s no way a doctor could have read all the latest research and publications at any given time.  The volume of information is too much for any human to achieve.  The idea is human and machine can work together here in a way that can help diagnose based on a large body of data in a rapid manner.  With some history and questions, the information that goes in is relatively small, the information that is evaluated against is vast, and the information that is returned is specialized and more importantly relevant.  The potential application could help diagnose, help patients choose the right kind of doctor, and more.  McQueeney contrasts this with traditional search engine technology.  What a system like Watson can do is not simply returned a response from an inverted word index, there is a computational leap in what is analyzed in a query.  In these scenarios, the response may not even have any of the words that were in the original query.  This leap in analysis against entire body or ‘corpus’ of information as McQueeney refers to it, is more significant and has a great cost.  All in all, a fascinating discussion that is worth watching.

Looking forward

In the past data was distributed in the construct where few people had a lot of knowledge and were able to share that with a lot of people.  Applications are increasingly possessing knowledge, changing that picture.  McQueeney illustrates that through the medical anecdote.  There is at any given time a common body of knowledge, based on research, history, journals, publications and the sort.  However for a particular doctor using such a system, there are individual insights that can affect the response to a query and that is based on the individual’s experience, changing the outcome.  The truth that a system has to embrace is subject to change at any given time and that is a monumental task.

The next big challenge in computing is likely along the axis of emulating human brain functions.  Starting with the goals of emulating strategy, deductive reasoning and pure logic such as used in the chess challenges, the evolution took them to multi-faceted ideas and subtle meaning, such as used in the Jeopardy challenge.  The next step would go into emotions and feeling.  That subtle context that feelings, tone, expression bring into analysis would be that type of next-step challenge.

Beyond the obvious medical and financial applications, there are a number of significant real world scenarios where computing can meet needs in new ways.  McQueeney discusses applications at the government level, where there are scenarios where a time-sensitive element enters the equation.  In the case of situations where property or life are concerned, timing becomes very critical.  Gaining the advantage of selecting a specialized set of information that is affected by time critical scenarios is one potential application.  McQueeney also discusses enterprise archiving and cybersecurity as a big data problem.