What is Ubicomp?

UbiComp 2007, the 9th International conference on UbiComp, took place this week in Innsbruck, Austria. Due to its size and interdisciplinary nature, its difficult to predict the “themes” of the conference in advance. Therefore, following my return from beautiful Innsbruck, I wish to report on what’sin the area of UbiComp based solely on my observations.

29 papers were accepted into the conference. Approximately one third of the papers were either trials or surveys. This category features field trials of existing work, such as the Whereabouts Clock (a family clock that abstracts members’ locations into the categories “School”, “Home”, “Work” and “Elsewhere”) and Roomba (the robotic vacuum cleaner), along with surveys into topics such as how families share their computers.

Of the remaining papers, just over a third were privacy-oriented, an increasingly hot topic in the field. The use of location in authentication and encryption had a particularly large presence. A quarter of them were “advanced sensors” that infer user activities by doing feature extraction and classification on raw sensor data. One fifth were special-purpose applications that took a specific domain, found a specific problem, and applied some basic UbiComp techniques to ease the problem. Ten percent were power management related, and another ten percent related to the display of context information to the user to aid in their task.

Of the two thirds of the papers that were not field trials or surveys, I was surprised to see that less than a third of them were based on the vision of mobile computing, connectedness, and services everywhere. Two thirds of these were primarily focused on privacy or security in this area.

Some of the application domains of UbiComp that I noticed during the conference are as follows:

Smart-home: entertainment, hygiene
Energy monitoring/saving
Modification of human behaviour
Health-care: medical, industrial, personal:exercise
Consumers and marketing
Social

 

Sun-dried tomatoes: not preserved?

Turns out that drying them out acts as a preservative (adding them to olive oil simply creates a softer, richer taste) and you can actually make your own in the oven!

“ Sun-dried Tomatoes are ripe tomatoes which are placed in the sun to remove most of the water content. Twenty pounds of fresh, ripe tomatoes will dry down to just one pound of sun-dried tomatoes. Sun dried tomatoes have the same nutritional value as the fresh tomatoes they are made from: they are high in Lycopene, antioxidants, vitamin C and low in sodium, fat, and calories.

Sun dried tomatoes were born in Italy as a way to store fresh tomatoes for the winter. Fresh tomatoes would be dried in the hot sun on the tile roofs as a way to preserve them for the cold months. Today they are still dried in the sun, but in much larger quantities and under strict quality controls.”

sun-dried-tomatoes.com

 

Machine intelligence: Realistic or irrational?

An old discussion cropped up at the IEEE SMC UK&RI 6th Conference on Cybernetic Systems today when the chair of the session, J. S. Quinton, posed the question, can we make intelligent machines? The question was then double-barrelled by asking whether we actually want to make machines intelligent.

The audience was almost evenly split on both topics. Our boat of discussion wasn’t about to sink. The first question quickly turned into being largely relative to each individuals definition of intelligence, which, as I gather, is often the route that the debate takes. One participant reckons that any living thing that can manipulate its environment is intelligent. Ants therefore are intelligent; we have already realised machine intelligence without knowing it through Swarm Intelligence. Another participant recognised that we had already realised machine intelligence in computer games—his definition was also satisfied. Why are we still researching the area? Apparently, intelligence seems to be that which we cannot make a machine do.

Others, including myself, weren’t so optimistic about the prospect of machine intelligence. What about conciousness or free will? Can we make machines concious? Our chair responded with a nice analogy. Conciousness is similar to the notion of “wetness”. It is a property of the molecule created by bonding two atoms of hydrodgen and one atom of oxegen, water. Neither of the separate molecules posess this property separately. The property of conciousness is similarly created by fusing cells together. Can we mimic conciousness in a machine then? Well, can we mimic “wetness” using digital circuitry?

With one cartridge smoking on the floor, we reloaded to fire the second shot: supposing we can achieve machine intelligence, do we really want it? Again, the audience was split. Some say “yes”. It can help us in dangerous situations and do all of the things that we don’t want to do. It allows us to “play God”. We could ask intellectual questions and get answers derived through processing power that we could never hope to equal.

If machines do everything we don’t want to do, what will we do? A few thousand people may create the intelligent machines, what will the other six billion do? Reproduce?! What if machine intelligence gets into the wrong hands (politicians), which it would, would we have a society where machines tell us what we can and cannot do? Suppose our defintion of intelligence includes the ability to replicate and evolve—will machines introduce a new type of evolution, and where will they stop? Will there be a conference far into the future consisting of a bunch of machines trying to answer where those bloody nuisances of humans ever disappeared to?

Something that wasn’t mentioned, but that came up during lunch, was responsibility. For a machine to be intelligent does it have to have responsibility? Do we even want to go there?

This was a surprisingly fun debate experienced by a Phd student who’s happy for now to keep machine intelligence at the “peripherum” of his research area.

 

The fallacy of the “steep learning curve”

While “googling” recently in order to find the best way in which to use the term “steep learning curve”, I came across a site which seems to complement a book entitled “Common Errors in English” by Paul Brians. The page details how the term “steep learning curve” to describe a difficult-to-master skill is, in fact, mathematical nonsense.

The reason for this is that a graph with time on the x-axis and effort on the y-axis would lead to a shallow curve given that a large effort is maintained over time. A steep curve, on the other hand, would result from a small amount of effort required leading to a much larger amount over time — a task that begins by being easy and quickly becomes more difficult.

 

Complex Adaptive Systems

Following my recent discovery of the field of complex adaptive systems (CAS), I attended a seminar given by Professor Doyne Farmer from the Santa Fe Institute. My goal was to investigate how the theory of these systems relates to, or can be applied to, that of context-aware adaptive systems.

The seminar was a basic discussion about the meaning and the properties of these systems. The goal was to define what exactly is a complex system and what properties do they entail. Some of the properties are as follows:

• Hierarchy
• Distributed Information Processing
• Interacting parts
• Emergent behaviour
• Evolution
• History, lock-in, path dependence

I wish to try to explain each of these properties and possibly relate them to context-aware adaptive systems.

In order to give structure to a CAS, it can be arranged as a hierarchy. The nodes on each level interact with each other and cause emergent nodes on the next level. This is usually as far as we understand or can attempt to model these systems. The nodes on level 2 and so on, however, can also interact with eachother in the same way. Such a system appears to be too complicated for us to comprehend. I intend to read a paper by Herb Simon entitled “The Architecture of Complexity” on this subject.

Distributed Information Processing is another property of CAS. Large tasks are delegated to smaller parts of the system that are specialised in more detailed tasks.

Components of the system may interact with eachother and as a result the emergent behaviour is not obvious from the parts alone.

Complex systems evolve over time according to a number of factors such as price of a product for example. If a price rises, more products are produced, and so on.

History plays an important part in CAS. The behaviour does not only depend on the state that the system is in, it also depends on the means by which that state was reached. This is known as hysteresis, a subject that I wish to become more familiar with.

Other keywords: Organisation, structure, function, ontogeny

Synomyms: cybernetics, self-organisation, plectics, emergence