Tuesday, October 21, 2014

Michael Jordan on the delusions of big data

Michael Jordan (the one from Berkeley, not the basketball player) gave an interesting interview to IEEE Spectrum.  it is recommended to read his own words. 

Some of the highlights of Jordan's opinions are:

  1. Using brain metaphors for computing is misleading:   computing does not work like the brain, this is also includes one of Jordan's expertise areas - neural nets.
  2. He says that the advances in computer vision lead us to be able to solve some kind of useful problems, but we are very far from giving machine the vision capabilities of a human
  3. "Big Data" is over-promising.  One can prove many false hypotheses using big data methods.  This is similar to building bridges without a theory of how to build bridges,  some may  survive, and some will collapse... 
  4. If he will have $1B to spend on research, he will invest in natural languages processing...

I think that it adds to some other observations about the overhype of "big data" (for example, see my posting on Noam Chomsky's opinion couple of years ago, or Tim Harford's recent article). 

Sunday, October 19, 2014

The Physical Web - by Google

Google recently revealed the "Physical Web" project.  This project is aimed at "interaction on demand" which will be a standard way that everybody will be able to consume data from devices connected to the Internet (AKA "Internet of Things") without the intervention of applications.   
This idea reminds of the idea of the grand challenge posed by the Event Processing Manifesto that was the result of the Dagstuhl seminar in 2010 and talked about "event fabric".     The "event fabric" challenge went further than get events on demand and also included processing event patterns on demand which I believe will be the next step to create access for everybody.  The ability to compose patterns on demand by everybody is a key to making this real-time data useful and complete the IoT revolution....  I am planned to give a talk related to this idea in early November in a workshop adjacent to the ACM Multimedia conference in Orlando... Will write more on this later...

Wednesday, October 8, 2014

On the history of STORM by Nathan Marz

Nathan Marz, the guy that is behind the Storm Apache incubator project.
Storm has definitely became the most common stream processing platform.  This year I am scheduled to teach a course about business intelligence, and my view of business intelligence includes the real-time business intelligence.  The students will practice Storm. 

Recently Nathan Marz wrote in his Blog about the history of Storm and lessons learned. 
I think it is worth reading... 

Tuesday, September 30, 2014

Some insights from the talk of Richard Soley in the IoT summit in INTEROP

The opening speaker in the IoT summit yesterday was Richard Soley, the person behind Object Management Group.  Richard talked about the "industrial Internet". He started his talk by having a nice slide in which the Internet now substitutes many thing we have done in the past
However, not everything changed by the Internet, since in many cases enterprises lack the "Internet thinking".  Examples are: manufacturing, energy grids, jet engines, oil and gas exploration and more are handled exactly as were done before.

This is due to the fact that the people involved including technical people are stuck in the way of thinking of the past.

Richard talked about the Industrial Internet Consortium which is a separate entity and not part of OMG (a correction that Richard made to my original posting). 
It has 85 members (at the time of the talk) and growing.  It is intended to study testbeds  in this area.  The Internet of Things is a crucial component in the industrial internet game.

One more insight from Richard is that "people don't read".  Everybody re-invents the wheel, since the current generation of professional people don't read and are not familiar with the state of the practice.  This is consistent with our finding in the event processing area where people prefer to reinvent the wheel and don't even know the wheel exists.  Another perspective of Richard's talk you can find in the article by Chris Taylor, who was the session organizer.   

Monday, September 29, 2014

My talk in INTEROP 2014

My talk on Internet of Things and Personalization is available on Slideshare.  This is a shorter variation (with some additions) of my DEBS'14 tutorial.  Enjoy!
I am also planned to deliver a variation of this talk twice more this week: in IBM Research at Yorktown Heights,  and in Stony Brooks University. 

I'll write more insights from the other talks - later. 

Friday, September 19, 2014

On the Proasense project

I have not written for a while... Busy days.  I'll write about my daily work in other opportunity.

Last week I've spent a couple of days in Athens, Greece, as a member of the advisory board of the Proasense project.  This project deals with proactive computing, which a paradigm that I have been advocating  for several years.  There are now couple of EU projects I know (the other one is Speedd) in the proactive space.  The proactive idea is that problems can be eliminated or mitigated before they happen.   The Proasense project employs two interesting use cases: one in the manufacturing area, and the other in the oil drilling area, an application that we have investigated a couple of years ago.

I am still in the opinion that proactive computing will be a major paradigm in the future, and will follow this project during the next couple of years with interest.   

Wednesday, August 27, 2014

On Fog Computing

Cisco that advocated the term "Internet of Everything" is now advocating the term "Fog Computing". these terms are related.    According to Cisco the cloud computing model is being replaced by the fog model.   
While in cloud computing all computation is done in a remote computing center, in fog computing the computation is distributed between local processing ("at the edge of the network") and remote processing. The relationship to the sensor world is straightforward.   A site might have multiple sensors. Some of the processing can be processed locally, and some need to be processed in a remote place, furthermore, this may be dynamic and tuned in real-time.   The picture above shows the before (cloud model) and after (fog model).   The example is energy system.  There may be processing done in a processor located in a single house which takes into considerations all sensors installed in the house.  There are other types of processing that related to data from multiple houses and need to be processed in a place where all data is available.  

Note that nothing is really new (besides the names).  Cloud computing is a new name for an old computing model that was once called "service bureau".  In the past the "cloud" was a single mainframe, and the edge where collection of dumb terminals.  Now the cloud is a grid, and the edge has processing power, but the principle is the same.    Fog is also an old principle of distributing the work reminding of N-tier middleware. 

I guess that the fog model is indeed more appropriate for IoT scenario than cloud model, in some of the projects that we are now planning within the Institute of Technological Empowerment,  are indeed fog based.  The sensors are going to be communicating with a local processor (which may be as simple as a tablet) with some processing done on the local processor, and some on the cloud.  This brings us back to the idea of event processing on mobile.   

After cloud and fog, we are waiting also for some - wind, rain, and sleet....