Causality vs. correlation - statistical reasoning is not enough - NY Times Interview with Dave Ferrucci

Dave Ferrucci, who was until several months ago an IBM Fellow  and was known as the father of Watson, was interviewed by the NY Times in his new working place at Bridgewater Associates.

In the interview Ferrruci somewhat continues the line of thought of Noam Chomsky,  saying that AI has concentrated around statistical reasoning based on correlations, but the drawback is that one cannot understand why the prediction made by the statistical reasoning is correct.  While Chomsky bluntly stated that statistical reasoning does not create a solid model of the universe, Ferruci claims that a complementary approach is required -  understanding causality.    This is a rather old issue, in symbolic logic, there is a distinction between "material implication"  which states that  IF A is true then B is true, and the meaning is that always when A is true then B is also true, which makes a sentence like  "If the week has seven days than  the capital city of France is Paris" - a valid statement in logic.    Entailment, on the other hand, said that "A ENTAILS B" if it is necessary and relevant, in other word, there is a causality among them.  Thus, Ferruci concentrates now on building causality models to model the world economy.      I concur with the assertion that understanding causalities give better abilities of reasoning and prediction.   As David Luckham already noted, causality among events is one of the major abstraction of event processing models.   Here is a rather old discussion about causality of events.  

On Artificial Intelligence

This morning, in the opening session of AAAI'11  there has been a panel of people who participated in the first instance of AAAI in 1980 who compared AI research then and today.    Hearing them I could not have seeing in my mind the picture of  the late Professor Yeshahayu Leibowitz, maybe the smartest Israeli of our time, who was known in his blunt controversial statements and opinions (e.g. he has been religious, but claimed that religion is just a set of axioms he taken upon himself and had nothing to do with reality).  The reason I remember him is because in the height of the Artificial Intelligence boom he wrote an article in one of the daily newspapers entitled  "Artificial Intelligence - an oxymoron", which claimed that intelligence is inherently a property of a living entity and cannot be artificial.   Since then I always addressed the term AI with a grain of salt.    AI research is very diversified, and there are sessions in this conference about: multi-agent systems, description logics, social networks, machine learning, natural language processing, search, knowledge representation and reasoning, planning, search engines, reasoning under uncertainty and more.    It also seems that like academic research, each paper concentrates on a narrow aspect of one of these areas,  it will probably take integration of all of these and more to create a real artificial intelligence that will break the Turing test.  It does not seem that this is a focus of current AI research, it actually broke away from the attempt to create artificial intelligence to solve specific goals.   Maybe the AI community needs a grand challenge.     The first keynote address was of Dave Ferrucci, the principal investigator of the Watson project who won the Jeopardy! game earlier this year.   It is probably the strength of industrial research that unlike academic research can gather multi-disciplinary people and focus them on a single goals.  Dave told us that in order not to stay focused on the goal, the researchers did not publish a single paper in four years, and started to publish all the scientific results only after the win - this is against the academic DNA, where people are measured on quantity of publications.     Tomorrow another day of AAAI,  starting the l-o-n-g trip home on Thursday.  I also visited the IBM Almaden Research Center yesterday, but will write about Alnaden in a separate post.

On Watson and Event processing

This is a picture taken in the DEBS conference last week, we had an invited talk by Eddie Epstein from the Watson team, and he ran a round of Jeopardy!, two of the conference participants against Watson, here you can see that Watson has -800 point (click on the picture to see it clearly), but it recovered and won.   We invited the talk on Watson, not because it somehow related to event-based systems, but because the conference took place in the Yorktown auditorium, where the famous Jeopardy! game in which Watson succeeded to beat two of the all time champions was recorded.


Today I've found a Blog posting by Shalin Shah from Vitria, with the promising title:  

IBM’s Watson: What Does Complex-Event Processing Mean For Customer Experience Management?

So I tried to understand what the author thinks is the relationship between the  two,  the answer according to the posting - both of them can be used for operational intelligence.

Indeed, there are now efforts within IBM Research, to determine what are the next steps, since in essence Watson is "deep question answering machine" there are some areas that seem to be killer applications of this technology, among them are: medical diagnosis and helpdesk/contact center in which agents need to answer questions in a lot of areas. There are some others as well.   

From technology point of view, Watson works in a different paradigm relative to event processing.  It is not event-driven, but is based on a knowledge stored in books, encyclopedias, and other sources.  What it does in real-time is - question understanding and question answering using statistical reasoning, and massive computational power.   

The interesting question is what can be a synergy between question answering machine and event processing,  here we can think of two sides:   an event processing system is being assisted in Watson-like system in order to determine contextual information that can be used for evaluating assertions, or classify events into context instances.   On the other hand a question answered can trigger event.  Or the question answering system can be monitored by an event processing system.     One can also think about real-time update of Watson's knowledge-base as a result of event, which is not the way Watson currently works.    I think that there are various more synergies between the two types of systems.

More - later.

Watson meets DEBS 2011

DEBS 2011 will take place in the Yorktown Heights auditorium, the same site in which the famous recording to the Jeopardy! programs in which the computer has won against two human champions.  The Watson hardware resides very close to this auditorium.  With some effort we succeeded to bring Eddie Epstein, one of the senior persons behind Watson has agreed to be an invited speaker in DEBS 2011,  Eddie will also provide a demo of the system.    


The full program of DEBS 2011 will be advertised around this weekend, the list of accepted papers can be found on this Blog.   


It should be note that if you plan to be in DEBS -- plan to stay for the entire conference, besides the 33 presentations, the attractions will be spread among the days  - no day can be skipped!


Monday, July 11 will be the  tutorial  day, and in the late afternoon there will be the EPTS award granting event. 
Tuesday, July 12 will have the two industrial keynote speakers: Chris Bird and Don Ferguson (see here for more about the keynote speakers),  and in the later afternoon a reception, and after getting some alcohol, the "gong show",  in which the participants can express outrageous ideas about future features and utilization of event processing technology.   
Wednesday, July 13 will have the keynote talk of Johnnes Gehrke,  the demo and poster session including the grand challenge,  and the conference banquet with artistic performance, and the DEBS awards (best paper, best idea in the gong show, best demo and maybe more). 
Thursday, July 14  will have the keynote talk of Calton Pu, and the invited talk of Eddie Epstein on Watson.


None of these day should be missed -- plan accordingly!


The fifth day - July 15 will be the PhD workshop adjacent to the conference.    


See you all in Yorktown.  

Watson - a machine beats human in Jeopardy!

One of the amazing achievements of IBM Research is the recent triumph of the "Watson" program, whose avatar  is seen above in the Jeopardy! game, beating two human champions.  This raises back the question whether computer thinks, and whether we are getting closer to the "singularity" vision between human and machine intelligence.     While this game is a question answering one,  a complete machine intelligence will also require the ability to detect events and contexts, and react within a changing environment.  I suggested (as a joke) that the next game that a humanoid robot should strive to is the "survivor", which requires other capabilities, like social capabilities, but I think that this challenge is somewhat beyond even the current wildest dreams.