Personalization - the next frontier

Back to DEBS 2014.  The last keynote speaker was Manish Gupta, who has been a colleague in IBM Research in the past, and now is is heading the Xerox research center in India.  Manish talked about personalization. He started the talk by saying that in the past everything was personalized - clothes, food, furniture. But only a few rich people could enjoy it.  The industrial revolution brought the mass production.

Indeed most products and services we consume are not tailored to our own individual needs, but rather standard on-the-shelf products.    We are now facing a departure from this mass production trend, and back into personalization.   Manish talked mainly about healthcare   Today, there is a tremendous progress in this area, enabling personalized treatment of some diseases, like personalized cancer treatment based on genetic patterns.  This is true in other areas of life.  We were exposed to personalized advertisement, but in most areas we are still in the mass production.   The current technology enables personalization by understanding individuals context on one hand, and classification of individuals on the other hand.  I'll write more about the technology about personalization, and how the combination of Internet of Things and personalization are changing the world, in subsequent posts. 

DEBS 2014 - the first two days

The picture was taken at the entrance to the computer science building in IIT Bombay where the DEBS 2014 conference is being held.  This is my first visit to India, and it is certainly different from anyplace else I've seen, I'll write about my impressions from India after I'll return.

The first day of the conference is the day of tutorials and PhD students section.   I have already written about my tutorial,  on the second half of the day Nenad Stojanovic delivered a tutorial about event processing on 

 mobile devices, following the start-up he is engaging in now. 

The second day started with the first keynote of Minos Garofalakis who talked about the work on reducing communication cost in stream processing by geometric reasoning.  I am familiar with this work that is done in collaboration with the Technion. Minos presented communication as the most scare resource that need to be optimized, but in answer to my question he admitted that this is only one of multiple factors that affect the performance.   

Other interesting talks were the one by Alessandro Margara about learning event processing rules.  The talk was interesting and his results look good,  however his work is restricted to very specific patterns and also to a derived event that is observable.   Many of the derived events are virtual, and the relationships between events that issue the pattern, and also the different dimensions of context, make it difficult in general to use machine learning techniques to automatically generate rules, and I think it is still largely an open issue.

Beate Ottenwälder gave a talk about reuse of patterns for optimization purpose.  This might be useful in some cases (where there are overlaps between patterns). 

 In the beginning of the day the organizers presented some statistics about countries from where the submissions sent from, and for accepted papers.   Not surprisingly, Germany is the largest country of activities in this area - both in submission and acceptance.  This year the acceptance rate was 9%, the lowest ever for the conference.  The organizers also presented some metrics about citations of the conference that are coming close to the major conferences and journals in the distributed computing area. 

More on DEBS 2014 - later.  

My talk in DEBS 2014 on the Internet of Everything

I am writing this post from the hotel "Meluha the Fern" in Mumbai.  Arrived here on Friday and had also an opportunity to do some sightseeing. Will write my impressions from Mumbai at a later phase.

Today DEBS 2014 started, the conference is being held in IIT Bombay.   The first day has been the tutorial day. I have delivered (by myself, my co-authors did not arrive) a tutorial on the "Internet of Everything".

This is the next in the tradition of tutorials that I am giving in DEBS since 2008.   As usual I have posted the tutorial on slideshare.  The problem with the slideshare conversion is that it messes up the animations, but I guess that it is readable anyway.  I'll write about the rest of the conference soon.  Enjoy!

Do we need Asimov's laws?

MIT Technology Review discusses to a recent article entitled "Do we need Asimov laws" by Ulrike Barthelmess and Ulrich Furbach from University Koblenz.    The Asimov laws celebrate now 50 years and this triggered some discussion.  For anybody who forgot (or never read Asimov - my most favorite author), the three laws of Asimov are: 

1. A robot may not injure a human being or, through inaction, allow a 

human being to come to harm. 

2. A robot must obey the orders given to it by human beings, except 

where such orders would conflict with the First Law. 

3. A robot must protect its own existence as long as such protection 

does not conflict with the First or Second Laws.

Actually in a later book Asimov made an exception and defined a zeroth law that puts the benefit of humanity above these laws.  

In the science fiction literature, a notable resistance to these laws are shown in the trilogy by Roger Mcbride Allen in the trilogy started with Caliban.

The article discusses that the three laws were dealing with different fears of people from the concept of robots, and asserts that the three laws were not implemented in reality neither in the autonomous vehicles projects nor in other robotic settings.   Furthermore there were other claims that Asimov's three laws cannot protect us.    Today robots also used for military purposes, and thus are by definition contradict Asimov's ideal about Robots as peace generators.       The authors set a moral principle: 

   "It is not allowed to build and to use robots which violate  Asimov’s first law!" 

Actually a counter opinion is that it is better to jeopardize a robot than to jeopardize a human in combat situation, the implementation of this moral law has nothing to do with robots, it has to do with the culture of settling disputes in violent way, this is what should be eliminated - but this is another story! 

PEW Research report on the Internet of Things in 2025

PEW Research center published this week a comprehensive report  entitled: "The Internet of Things will strive by 2025".   This report is based on a survey that 1600 responders participate in.

I have copied the Table of Contents of that report:

I am now working on a three hours tutorials on Internet of Everything that I plan to deliver in DEBS 2014 in Mumbai. 

On the shareholder value myth

I came across a small booklet entitled "The Shareholder value myth" by Lynn Stout.   I now have the privilege to look at the corporate world now from the outside so I can write more freely from my academic seat now.   The book starts with the assertion that during the last 20 years there is a common practice to see the "shareholder value" as the only consideration in corporates' strategy.  Actually when I studied in an MBA program in the early eighties,  we learned that there are multiple stakeholders: shareholders, employees, customers, suppliers, regulators and the social environment, and the management needs to balance the interests.    Since that time the equation has changed to have shareholder as the primary stakeholder. The main thesis of this book is that this is actually a wrong thing to do, since it biases towards short term (the speculative shareholder vs. the long-term shareholder).  Since the author is a professor of law, she starts by legal claims that disputes the common thinking that the shareholders own the corporate, and discusses the relations between these two entities.  She claims that the current economy takes it to extreme by saying that the goal of maximizing the value for shareholders is a goal the justifies all means to achieve it. Stout claims that this kind of thinking is typical for psychopaths, and that most people, thus most shareholders are not psychopaths.

The consequences according to Stout are damaging to corporates, employees, customers and society.  The drive for short-term results, fueled by linking senior executives compensation to short term goals, and cutback of R&D, employee benefits, and quality and affordability for customers.    It also triggers unethical behavior.  The good news according to Stout is that we see first signs that this paradigm starts to decline, however the "Wall Street" culture is still quite pervasive. 

I'll write more opinion about socio-economical issues in time -- still learning it! 

Internet of Things - what's holding us back?

InformationWeek published an article this week by Chris Murphy entitled: "Internet Of Things: What's Holding Us Back".   In this article Murphy describes several reasons that hold us back from exploiting the potential of the IoT.  The reasons he mentions are:

1. The data is not good enough:  the claim is that the conception that all requested data is readily available is not consistent with reality, where data suffers from quality,  frequency and spatial coverage of the sensors, and data integration issues.
2. Networks aren't ubiquitous:   The product owners don't have control over the availability of networks
3. Integration is tougher than analysis:  The main problem is not to analyze the data, but to integrate all data needed for analysis
4. More sensor innovation needed: The stated areas of required innovation are - combine video sources which today are under-utilized; more-refined and more-affordable environmental sensors; software-defined sensor,a combination of multiple sensors plus computing power that sits out on a network and "calculates rather than measures."
5. Status quo security doesn't cut it.  Security systems for IoT should be radically different than those developed for traditional IP.

I agree that all of these contribute in one way or another to the difficulties around exploiting the potential of IoT.    Dealing with inexact or uncertain data is a major issue, a link to our tutorial on the topic can be obtained from this blog post.  What Murphy refers as "software defined sensor", is in fact, the ability to use multiple sensors and get sense out of it in real-time,  this is exactly what the event processing discipline produces, furthermore, our work on event modeling contributes to make it simpler. 

I am planned to deliver a tutorial on "Internet of Everything" in DEBS 2014 in Mumbai, where I'll discuss all these issues.  

More - later.