Image via Wikipedia

Howoldareyou.net presents a game to look at the pictures and try to accurately guess the person’s age. If you try it, you will see how (un-)precise our natural neural network model can tell the age of a person.
I wondered if there is somewhere a (artificial) data mining model that can outperform human estimation power. (And indeed, some people already gave it a try : Xiaodan Zhuang, Xi Zhou, Mark Hasegawa-Johnson, and Thomas Huang)

It first should be able to detect the face in the picture. That problem is already solved by Pittsburg Pattern Recognition.

Second step is to estimate the age. Which variables should you use ? Must be an interesting problem to solve. Perhaps when I have the time somewhere in the future ?

A similar problem is given by hotornot, but this must be a thougher one, because here it is not only the face that matters. And second what one person finds is hot, is not necessarily the same for another person. In the age estimation game the target variabele is something there is no discussion about.

However, a person can look older or younger than his or her real age.
Interesting point here is that the “intelligence of the crowd” : the average estimation should exactly give the apparent age.

Related articles by Zemanta

• Captcha cracking in JavaScript with Canvas and neural nets

                                     

In september of previous year a new phenomenon has entered the blogworld : Alphainventions, a website made by Cheru Jackson.

What is it ?
It is comparable to twitterfall : Alphainventions continuously shows blogs. You can let them come and go, or you can pause them. But how does Alphainventions decide which blogs to show ?
There are two very distinct functionalities intermixed

1. show new or updated blog posts

new blogs are added to the reading cycle so that they show up in Alphainventions. That is why people were surprised suddenly to see traffic to their blogs coming from alphainventions.

2. blog promotion (for free)

On Alphainventions you have also the opportunity of entering your blog url into the reading cycle, taking the place of another blog that is kicked out.

I think this mixing of two different functionalities definitely is no good thing.

What should it be ?

The alphainventions fellow should clearly separate the two : He should make on the one hand something like a “blogfall” (waterfall of blogs) that only shows the new blog posts and on the other hand a blogpromotion site.
For the blogfall, visitors should have the opportunity to choose between two view modes : 1)seeing the blogs themselves, as now is the case or 2)just see a list of blog names plus the titles of the figuring post so that you have to click on an item to see the blog post (preferable in a new window.
For both sites there should definitely be a selection algorithm where you can enter your selection criteria, your topics you want to see blogs from.

In any case Alphainventions is an interesting idea. But there is much room left for evolution, certainly much more than the few suggestions I made here.

                                     

Here are some email tricks, either to save some time, or just to fool the others …

1. send a twittermail = email not exceeding 140 characters (Jennifer Leggio)
2. write a short message in the subject, nothing in the email body
3. change your desktop email signature to “Sent from iPhone” – then write 1 sentence (
Kevin Rose)
4. change your desktop email signature to “Sent from twitter” – then write not more than 140 characters.

                                     

Image via Wikipedia

In Snippets, John Tropea quotes Clay Shirky who argues that the much discussed information overload does not exist. Normally filters take care of that. According to Clay, people who complain about information overload (are old and) do not use the available filters (like Digg, del.icio.us or google reader) or not properly.

I will not argue against his point of view, but I believe that the filters have to be much easier to set up. I expect this to happen with Web 3.0, the so-called semantic web. With a public profile, when you open your browser you will just have to ask for news, and the semantic web will filter it for you. No setup to do any more, exept perhaps at the beginning, where you will have to indicate what you want to see and what not. And that also will be limited, because clever data mining models will soon guess your overall profile and give you anything you normally should like. All this news will of cause properly be hierachised.
According to what you read and skip, your profile will continuously kept up to date.

Would that not be nice ?

Did you liked this post ? Then you might be interested in the following :
Web 5.0 : the telepathic web
KM 1.0 KM 2.0 KM 3.0 …
Piqqem : Prediction market for prediction errors
Web2.0 and the lack of process
End of privacy

Related articles by Zemanta

• Filter Failure
• 10 tools for listening in social media
• [Insite] Semantic search – an Interview with Brooke Aker

                                     

These days email tends to eat up a great deal of our working time. A lot of articles on the internet deal with being a slave of email, email overflow, or on how to cope with too much email.

As a reaction to all this email came the initiatives of Luis Suarez : giving up work email, where he tried to get the weekly number of incoming emails down to zero, or the email brevity challenge of Jennifer Leggio who encourages everybody to limit the email body to the length of a tweet : 140 characters.

All this email means only one thing : enterprises 2.0 are still rare ! If you work in an enterprise with no social tools, email is all you have got. And as the saying goes : “if you only have a hammer, everything looks like a nail”.
But it is hard to get an enterprise from 1.0 to 2.0. You have to break down some silos, as is discribed in this article from Connectbeam.

Anyway, the enterprise I work in is still in its 1.0 phase. Everybody only uses emal, phones and, yes,can you believe, still a lot of printed paper to distribute reports !!

So this week I realized that in our company there exists a little tool that permits instant messaging. It is not as good as, say, MSN, but for short instant messages it does the job.
As only a few are aware of the existence of this tool, you can imagine how my colleagues are surprised receiving instant messages in stead of short emails. And I hope it is contagious.

In the meantime I began to clean up my inbox an “sent” folder and also got to arranging my “knowledge” in a personal wiki (wiki-on-a-stick). My ultimate goal is, when I leave the office at the end of the day, to have my inbox empty !
I wonder if I will keep up the discipline ! But I know it is worth the effort, since everything will be so much easier afterwards.
What I am already accomplishing is not to let it grow any more. An important first step. And from time to time a little cleanup ensures that my email heap diminishes.

                                     

Related articles by Zemanta

• Before There Was Twitter, There Was Dave Winer’s Instant Outliner

Image via Wikipedia

When you attend a data mining course, even from one of the leading data mining software vendors (like SPSS, SAS, Salford Systems) you learn how to push the buttons fo the tool and the basics of the algorithms. An advanced course consist of more advanced algorithms like neural nets, or SVM’s.
But one of the oldest algorithms is decision trees. And yes, in the advanced course you learn briefly the new ways of working with decision trees, like bagging or boosting.
But there is more.
In this post, the drawbacks and disadvantages of decision trees will show you why I love working with them and in a second part I will show you the secrets of how to get better predictions from decision tree models.

Disadvantages :
– overfitting (but you can control it and use it to enhance the model quality !)

Advantages :
– readability : simple if then statements
– simple algorithm
– robustness : preprocessing and cleansing is an cumbersome job, but decision trees work well without it !!
– flexibility : the settings allow you to build from very small to very large trees
– overfitting : if proporly controlled
– simple powerful settings
– large models (exponential growth with deeper models) that give very “detailed” results, see next post.
– weak learner, so there is a lot of gain in ensemble techniques like bagging.

Getting better predictions.

BAGGING !
In numerous studies it is shown that bagging (bootstrap averaging) dramatically improves the quality of decision tree models (DTREG,Kristína Machová, František Barčák, Peter Bednár, Breiman ) Why ? decision trees are weak learners. This means that with different samples from the same dataset you can easily obtain very different models ( but of equal quality). By averaging the results of several decision tree models, you obtain an average model. What does this mean ? You can compare it to the mean, the standard deviation and the standard error of a normal distribution. If we see the best possible model as the mean, than you can look at one single decision tree as an observation with a deviation from the mean with a standard deviation distribution (about 1 chance out of three that it is more than one standard deviation away from the mean). With bagging the average model has one chance out of three to be more than one standard error away from the mean. One standard error equals one standard deviation devided by the square root of the number of trees, so one standard error is much smaller than a standard deviation, or : one bagging model is much better than one tree model.

But : you not only need a weak learner, you also need overfitting !
What is overfitting ? Normally it is described as “the model learns the noise”. According to wikipedia : the model has too much complexity as compared to the amount of data available. But since no model is 100% perfect, there is allways some noise, and hence each and every model is somehow overfit. It is the degree of overfitting that is all about. With decision trees it is very simple to determine this degree of overfitting by controlling the size of the tree.
What are the possible settings in this respect. You can define the number of layers (dept), the minimal terminal nodesize and the minimal splitsize.
Forget about the number of layers and set this allways as big as possible. It should never be a stopping criterion if you want a good quality model. (Only if you want a simple, small, readable model, then you can use this, but then it only serves for documentation purposes).

With the minimal terminal nodesize and splitsize you control the degree of overfitting. Set them too large and you obtain small trees with little overfitting but poor quality. Set them too small and you obtain huge trees with too much overfitting. It is necessary to do some experimentation and test the quality against a hold-out dataset. A good way to assess overfitting is to look at how the lift of your bagging model behaves for several selection sizes. Normally the lift increases as your selections are smaller. If at a given point the lift starts to decrease with smaller selections, or become erratic, you have too much overfitting.
Ideally, your lift versus selection size plot looks like this (not real data, but very much like my real data).

So why do you need overfitting ?
With bigger trees you have a larger degree of overfitting. But with bagging, you do not only average the real predictive value of the model, but also the overfitting part. And because it is overfitting, per definition it is random noise that tends to zero als the number of trees becomes larger. So you have to tune your settings that way that the overfitting is gone by averaging and the real patterns remain.
Good luck !

In my next post I intend to write about : How good can a model be ?

Other posts you might enjoy reading :
Data mining for marketing campaigns : interpretation of lift
Howmany inputs do data miners need ?
Oversampling or undersampling ?
data mining with decision trees : what they never tell you
The top-10 data mining mistakes

                                     

Related articles by Zemanta

• Oversampling or undersampling ?

Image via Wikipedia

The question in the title of this post is simple enough. The answer is not.
I believe selecting the right partner is finding a the ideal point in a three-dimensional space. And it is important to note that first this three-dimensional space is not our spatial dimensions of lenght, width and height, second that these three dimensions are probably a crude oversimplification of reality and third that all that follows in this post is written from the point of view evoluion of behaviour (in a darwinian or post-darwinian or later … meaning).

The three dimensions :
1) male fitness (as observed by females)
2) female fitness (as observed by males)
3) selection purpose

Male fitness

In a herd, flock, group, community or whatever contains some individuals, at least in social animals there is some sort of hierarchy. The same holds for humans. The alpha-male is the lucky guy who gets every girl or women he wants. The omega-male is the one who allways goes alone. And between the two extremes every value is possible.

Female fitness.

Is very much like male fitness : The alpha-female is the one men fight for (not necesarily literally), the omega-female is the girl nobody wants.

Selection purpose.

As I wrote in a previous post females select two sets of characteristics in men : first set: he must be as “alpha” as possible (strong, healthy, rich, social … ), whatever characteristics a women wants her children to have. Second set: he must be a good father. And when she does not find the two groups of characteristics in one single man, then many seek the first set in one man and the second in another (they do not need to know eachother …).
Something similar holds for the partner choice of men. A man can select a women to give her the first set of characteristics. A one night stand is sometimes all it takes to give her the right set of genes for her child. Or a men can seek a lifetime partner to raise his children and provide himself with a real home and family.

In the beginning of this post I spoke of finding the ideal point.
A very simple approach would be for a man : just go for the alpha-woman. But it is not that simple. There are right choices and not-so-right choices.
Let me illustrate with a simple graph of the first two dimensions.

The red diagonal represents a totally balanced selection : in the upper right corner we find the alpha-men, selecting the alpa-women. In the lower left corner just the opposite: omega-men selecting omega-women.

The green area is the easy area : It is where a men selects a women who ranks lower on the alpa-omega scale. It is an easy victory for the man, and the woman is happy because she has a better than expected man.
I think this is the more dangerous zone for the women, because chances are high that the relationship will be short. This could work for both in case they somehow agree that the selection purpose (dimension three) is more on the gene (sperm) transmitting side than on the family side.

The read area is more dangerous for the men. He can try to start a relationship with a women who ranks better than himself, but his chances will be be low and then he will find other men ranking higher and even risking agressions of whatever kind. Besides, he risks to be chosen, not for his “alpha-level” but for his family level, by a woman who goes to another man to obtain the right genes for her childeren.

So it remains a very complicated choice, the more that all this happens to a large extend subconciously. A very fertile soil for lovesongs !

Related articles :

Relationships/How Men Select Women
Beauty and evolution
How to select a mate

Related articles by Zemanta

• “And When You Give Them Your Car Keys, And Your ATM Card”

                                     

Image via Wikipedia

There is a lot written about the benefits of web 2.0 inside the enterprise but for this post I looked up some problems.

The first one is IT, and more specifically : security. They tend to build as many walls as possible to keep hackers, viruses and other malicious softwares outside. The problem is that they forget that nobody can get out any more !
I often tend to say that it is easier to stay informed about what happens in the outside world that what happens in my own department (we do not have anything web 2.0-alike !)

Second problem : If IT does not do it, the end users will do it themselves creating security problems.

Problem 3 : a lot of people believe enterprise 2.0 is critical to success, but also don’t really know what it is. A bit like pornography: they’ll pay too much, get over-excited after tiny results, but soon regret it (John Bordeaux)

Often people think it is about tools ! but it is not ! You have to solve a problem, instead of introducing tools.

Problem 4 : How to get people participating ? One of the difficulties about social software is to get people to use it. Bizar, since it is made for them. The internet is full of web 2.0 people (facebook, twitter, anything) but when you start something like that in the enterprise you have the hardest time to get people involved … unless you clearly create benefits for them. It must not be something they have to do besides their job, it has to ease their job !

And last but not least : hierarchy ! The bosses are afraid to loose their grip on the situation

                                     

Related articles by Zemanta

• IT chiefs urged to embrace enterprise 2.0

Image via CrunchBase

Jennifer Leggio launched a 2009 email brevity challenge : just limit your emails to twitter length (=< 140 characters).

And while you are limiting : simply put what you have to say in the subject line, and do not use the email body.
That way you can be twittering with email !

Enjoyed this post ? Then you might also be interested in the following :
A bunch of tools for twitter
8 web 2.0 predictions

                                     

Image via Wikipedia

Apparently, wikipedia is not a termite hill after all.

You know, the intelligence of the swarm, where a whole bunch of dummy individuals together create a masterpiece like a termite hill or a beehive.

In our common knowledge wikipedia is created by a “swarm ” of thousends and thousends of contributors. Kevin Kelly described this as follows :”Wikipedia is not the only hive mind out there. There’s the grand web itself“, before two sentences later he admits that “In fact a close inspection of Wikipedia’s process reveals that it has an elite at its center, … there is far more deliberate top-down design management going on than first appears”

Now Aaron Swartz conducted a study to find out who exactly creates the content of wikipedia.

It is striking that 50% of the edits are done by 0.7% of the contributors and 73.4% is done by 2% or 1,400 people.

So you can say that it is a very skewed swarm : The bulk of the work is done by a team of experts whereas a lot of the input, but only 26.6% of the work, is provided by the remaining 98% or approximately 68,600 people.

It is like a beehive that has in stead of one queen a huge royal family at the commands.
Or stated otherwise : it is very much like how they wrote traditional encyclopdia’s on paper.

                                     

Related articles by Zemanta

• Swarm intelligence and hierarchies
• Wikipedia proves its popularity with fundraising surge
• Who The Hell Writes Wikipedia, Anyway?

I did a little search for what people think what will be the trends for 2009.

An overall result is simply (and dull) : more of the same, meaning that people think that what is rising will continue to rise.

Number one : more social media, with as results :
everyone will become marketeers
and organisations, enterprises will get more social

there will be still more advertising on internet
there will be more virus attacks towards websites
advertising will become more multicultural

Customer service becomes more important
netbooks will keep on rising

And, of cause, Google will buy twitter.

Here is the list of my sources :

trendspotting
toptrends
scdsource
businessweek
intelligent measurement
trendwatching
vignette
Ross Dawson
Emarketer
Econsultancy
Trendhunter

:: :: :: :: :: :: :: :: :: :: ::

In a presentation on Slideshare, Dr Sven F. Crone of the Lancaster Center for Forecasting put his finger right on the wound when he talks about the myth of the best algorithm (Note : his talk begins at slide 116 !).
With real life examples he draws attention to the fact that the preprocessing of the data and the method of sampling is much more decisive for the quality of the resulting data mining model than the modeling algorithm.
I totally agree with him on this. But I want to do here is comment on his conclusion that oversampling is allways better than undersampling.
He IS right in HIS examples.

What happens when you do oversampling or undersampling ?

Oversampling : you duplicate the observations of the minority class to obtain a balanced dataset.
Undersampling : you drop observations of the majority class to obtain a balanced dataset, see illustration.

As far as the illustration goes, it is perfectly understandable that oversampling is better, because you keep all the information in the training dataset. With undersampling you drop a lot of information. Even if this dropped information belongs to the majority class, it is usefull information for a modeling algorithm.

But nowadays in big enterprises first there is plenty of data and second the data mining algorithms/softwares/hardwares are often limited in the amount of data they can analyse.
Ever tried training a model with a training dataset of 50Gb ?
So if you have that amount of data, maybe undersampling still leaves you with too much data and you have to use only a fraction of it.
In that case undersampling is better, oversampling is useless.

And one last remark: I do not believe that simple oversampling is a good idea, even with a largely unbalanced dataset. In that case you should choose a modeling algorithm that can handle that imbalance, like for example a decision tree.

Did you liked this post ? Then you might be interested in the following :
The top-10 data mining mistakes
Text mining : Reading at Random
data mining with decision trees : what they never tell you
Toddlers are data miners

:: :: :: :: :: :: :: :: :: :: ::

I want to thank you all for visiting my blog. After all, that is the only reason why I keep writing.

The end of 2008, my first blogging year.
Although I do not allways post as often as I would want to, I find I spend a bit too much time on my computer. Quite a contradiction !

Nevertheless, here are some statistics of my blog, showing that I have been quite busy this year.

• number of posts : 170
• number of views : 6,852
• number of comments : 80

And here a short list of the most often viewed posts this year :

• Can/do/did dragons exist ?
• The differences between animals and humans
• 15 Knowledge management failure reasons
• top-10 lists on evolution
• Wiki on a stick (3)
• Ant Colony Optimisation : Ants or Locusts
• Ant behaviour datamining: Complete algorithm
• did dragons exist ?
• Do Stock Traders show Swarm Intelligence
• Top-10 lists on Knowledge management

If you have any comments, suggestions, preferred categories, feel free to write a comment or to send me an e-mail. I will definitely keep your whishes in my mind next year.

Happy new year and may everything in 2009 be better for you than in 2008 !

Zyxo

:: :: :: :: :: :: :: :: :: :: ::

Image via Wikipedia

Web 2.0 is still going on and a long way before being replaced by Web 3.0
Web 3.0 : everyone is still talking about, nothing much happens yet. Still some years before we will see something decent, another 5 years of maturing before the next release will be on the study desk :
Web 4.0 is stil a great question mark, but something will show up. It will take time, but by then we will be the year 2020. Say 2030 before we will have a mature web 4.0.
Remember that people still need the time to early adopt, become mainstream with the first applications and than to get ready for the next upgrade.
And in the meantime : brain implants, helmets … to read/write directly into your brain will get better and better, smaller and smaller, cheaper and cheaper.
And remember : any electronical device that can read your brain can be connected to a wireles modem to connect you to the internet … to connect you to anyone ore anything you like : your wife, your facebook friends, wikipedia, your radio- or TV station, your coffee machine…
ThinK of chatting online just by thinking to one another.
Think of recording your dreams and watch them the morning after or sharing them, say on youtube …
Think of calling your dog which is three miles away …

Related articles by Zemanta

• Brain Electrode Implants Helps Paralysed Monkey Regain Arm Movement

Other related articles by Zyxo:
Telepathy via Circuitry: A New View of the Internet (Steven J. Searle)
Our Telepathic Future (Dave)

                                     

Image via Wikipedia

The versions of Knowledge management go hand in hand with the versions of the internet : 1.0, 2.0, 3.0, …

But how far are we with knowledge management?

In an interesting presentation, David Gurteen gives the most important characteristics of Knowledge Management 1.0 and 2.0. In the table I summarise the highlights of it.

Knowledge management 1.0	Knowledge management 2.0
techno-centric	people-centric
command and control	social
centralised monolithic systems	decentralised ecosystems
email, newsletters, databases	social tools (blogs, wikis, IM‘s)
KM is extra work	KM is part of my work
IT selects the tools	I select my tools

 
These are the same differences between web 1.0 and the social web.
But what about the semantic web ? Is there also a Semantic Knowledge Management ? (or KM 3.0 ?)

I assume there not one yet. But is there one to come ? In any case The 1st Workshop on Knowledge Management with Web 3.0 has yet to be hold.

In an article on IEEE Distributed Systems Online about Semantic-Web-Based Knowledge Management John Davies, Miltiadis Lytras and Amit P. Sheth summarise their view on the topic. The most pertinent topics are :

• metadata management and extraction of attribute values for improved search
• integration of knowledge creation and use, integration of technical and human intelligence
• using semantic web technologies to exploit information sources like wikipedia

But this all has still a long way to go.

Related articles by Zemanta

• The Semantic Web Pitch by Nova Spivack

Image via Wikipedia

An article in the economist describes three hypothesis of why something like music has come to existence.
They discuss three possible reasons :

1. it helps for partner selection, just like for example the peacock tail does
2. it has a social function to bind groups of people together which gives them an advantage over less binded groups
3. it has been accidentially invented, and offers no advantage

It is clear that the third is contradictory to the former two AND it is not very likely that something as important and ubiquitous as music could have come into existence without any advantage for the performers.

Image via Wikipedia

Very interesting, intruiging and loaded discussion about science versus religion on the blog of Daniel Florien : Science is limited by its refusal to make stuff up

In a comment of Vorjack the phrase “If something doesn’t have an effect on the world, then it doesn’t matter as far as science is concerned. So anything that effects the natural world is natural. The supernatural, if it exists, is meaningless to the discussion.” makes me question : has God an effect on the world ?

Sure he does ! Nearly every war was/is powered by religious beliefs.

Although god himself is not tangible, the concept is there. But when was he born ?

Firstly : Men allways had fear for the unknown powers.
Secondly, men organised themselves in tribes, villages cities, nations … : together they were stronger. And someone had to be in charge to keep the community together.
The one in charge deserved respect (or fear) (otherwise he could not stay in charge), which was somehow correlated with the size of the community.

For a long time the two coëxisted : 1) natural forces, like sun or lightning who were seen as the gods, and 2) the human chief, king, emperor.
Until the roman emperor decided he was god also and had to be respected accordingly.
But although this was a bit over the top, the click in the minds was made : the natural forces were not the gods, and neither the human emperors were god, but the god was something like a human, someone who was not afraid of the natural forces, who even was in control of those scary natural forces, so much powerful than men.

So the intangible, supernatural god was born, evolved from the natural ones.
And even if he only exists as a bunch of memes in some human heads, memes evolve too.

Related articles by Zemanta

• World Wide Web spreads like religions

Image via Wikipedia

Nothing really new on the singularity summit 2008.

But it still is a bit weird to see all of these things together and wonder : where will we stand say in 25 years. If you look at what happened the last 25 years. It still goes faster and faster. Last year Facebook was something for early adopters, nowadays everyone is on facebook and MSN is forgotten.
Nanomachines, electonic chips implanted in our brains, baby on its way, selected on not having a cancer gene, google being the first to see flue outbreaks, solar energy production doubling each year, advanced robots, cell reprogramming, etcetera.
It is really a fascinating time !

Image via Wikipedia

According to wikipedia, we can speak of intelligent design when “certain features of the universe and of living things are best explained by an intelligent cause, not an undirected process such as natural selection“.

The most important part here is : not an undirected process. In other words : Intelligent design is, well, a design made by something or someone intelligent, meaning that the outcome is wanted from the beginning.

OK. But what is natural selection ? Let us turn again to wikipedia : “the process by which favorable heritable traits become more common in successive generations of a population of reproducing organisms, and unfavorable heritable traits become less common, due to differential reproduction of genotypes“.
Apart from the fact that the words “favorable” and “unfavorable” have no meaning whatsoever in this definition, it is clear that the result of natural selection is evolution : some heritable traits are more common, others are less common in successive generations, meaning that the population changes, evolves.

But one typical characteristic of evolution is that if favours complexity. From simple molecular complexes to simple animals to complex animals, to intelligent animals, to animals that select other animals and influence, or rather design, the direction of these other animals’ evolution.
So intelligent design is just another level of evolution.

Enjoyed this post ? Then you might be interested by the following :
Web 5.0 : the telepathic web
Do Stock Traders show Swarm Intelligence?
Swarm versus intelligence
Piqqem : Prediction market for prediction errors
swarm-information-transfer-techniques

See also this interesting article from Philip Dorrell

Image via Wikipedia

What will we become ?

A recent Scientific American article discusses the idea and apparently has attracted the interest of a lot of people.

Here are some of my own thoughts about the subject.

First : about evolution itself :
Evolution = change + selection.
.. Change by mutation, crossing over, mixing of father and mother genes
.. Selection : forced by the environment and strongest in changing environments.

It is clear that evolution never stops. There is no reason why mutations and crossing over would quit their jobs. And with the globalisation there has never been more mixing of different father and mother genes.

The habitat where the human species dwells has never changed at the present pace. Every generation sees another mother earth. So it is clear that selection is on its full speed.

OK, let us assume that the human species continues evolving. What is the direction ?

Alternatives are :

• bigger humans. According to Olivier Curry : human species is set to reach its peak in the year 3000, growing taller and living longer thanks to improved nutrition, lifestyles and increased medical knowledge. They will also modify themselves – through technology or otherwise – to attract partners and will therefore be better looking. ‘Race’ will also be a thing of the past – by the year 3000 all humans will have ‘coffee’ coloured skin. As far as I am concerned this is complete bullshit. 3000 years is far to way ahead to make predictions, and if people live that long they will probably intelligent enough to make smaller humans, with tiny ecological footprints.
• bigger brains. Not quite probable. If you see the increase in intelligence of humans during the last centuries, or even decades, it is clear that our brains are big enough. And more : to get bigger brains would mean that people with the biggest brains on the average have more children. And that I do not believe.
• Connected brains. Absolutely sure. It is the most normal thing that all machines become connected and exchange information. The connection of human brains to machines has also started (see for example here and here). So it is only a matter of time that individual minds will communicate with one another via our electronical gadgets. Real telepathy !
  According to Kevin Kelly’s post what comes after minds ? there will be a biosphere of minds, an ecological network of many minds and many types of minds that would have its own meta-level behavior and consequences.
  I think that meta-level is still relatively far away, unless the next alternative is there fast.
• Machines ? If people continue to develop thinking things, like for example Asimo robots brainpowered by biological brains, it could be that the artificial brainpower will quickly surpass our own and what will happen next will be too complicated for a human being to understand

And two more things left :

Group selection : the selection of the best tribes will not exist any more : globalisation has left only one single (huge) group. To select the best, you need alternatives, which is impossible if there is only one left.

If people get smart and technically savvy enough to do the selection themselves : who will decide what to (de-)select ? (Kristin Jenkins on serendip : “Frighteningly, it may be” (Conklin, 1922))

Image via Wikipedia

A recent study by Gunter Wagner and other researchers at Yale and Washington University show that higher organisms do not have a “cost of complexity” — or slowdown in the evolution of complex traits.

Cost in evolution : because of the complexity, the effect of a single mutation is diluted and has a smaller impact.
Benefits : one single mutation has often impact on more than one trait.
The benefits make up for the costs.

I see some correspondence with complex data mining models (from for example bagging, random forest or artificial neural network algorithms).
The cost in complex data mining models is the burden not only of training the model but also the burden of extracting all the variables that are used by the model (sometimes one of my models uses more than 100 variables). And the beneficial effect of adding new variables to an allready complex model generally is relatively small.
The advantage is the dilution : when one variable contains errors or is missing for one client, the other variables together with the complexity of the model take over and the model still comes up with a good, usable score.
So in stead of wasting my time to get all variables right and to clean up all the data, I just throw the whole bunch of variables in the algorithm and let the computer do the dirty work. It usually comes up with very good and very robust models.
I love the data mining saying : “more is better” and by that I mean more observations and more variables.

Sas posted a commercial video on youtube where Santa Claus and his co-workers share their exitement over the advantages of using SAS in the Santa Claus enterprise.
At least Sas grabs the opportunity of the year, but I still find it a bit cheap and the content of the video is to serious to be funny. A rather boring commercial talk in red and white.

At Fastcompany I found what 8 experts predict for 2009 about web 2.0.

I summarize their predictions :

• stronger identity portability
• more user control over their own data
• more company revenue in social space
• marketing shift towards interaction
• exchange of profiles, communities, between social networks
• social media becoming more accessible, outside computers
• more location-aware services
• better aggregation services

In a short sentence : more and better web 2.0

Image via Wikipedia

What is evolution ?

Let us start with a wikipedia definition : “In biology, evolution is a change in the inherited traits of a population from one generation to the next. ”

Yes, “in biology”. But there is anoter one : Universal evolution, parallelly introduced by Vladimir Ivanovich Vernadsky and Pierre Teilhard de Chardin : They describe nine levels : 1=strings, 2=elementary particles, 3=atomic particles, 4=atomic, 5=molecules, 6=eobionts;7=protozoa,8=metazoa,9=socialisation, where levels 6 to 8 represent the biological evolution.

Recently there was also a paper about evolution of minerals, which I already mentioned in a previous post, and then there is also something like artificial evolution as discribed in Chapter 6 of “Artificial Life” by Prof. Dr. R. Pfeifer, Dale Thomas and Max Lungarella.

So there is evolution everywhere. But is it the same evolution ? I like the things simple. So let us come to the very basic idea of evolution, and for that I will use an extremely simple example.
Say we have two experimental setups, eacht with 100 red balls and 100 blue balls.
After a certain time, in the first setup, 25 of the red balls have changed colour and turned blue (do not ask me why, just accept it). At the same time 25 of the blue balls have turned red.
In the second setup only 25 red balls changed into blue, all blue balls stayed blue.

Now the very simple question : in which setup did we notice evolution ? I would say : in the first one. Why? Because the group of balls as a whole changed. In the second one, although some individual ballc changed, the group of balls remained the same.

So, if I can permit to write a new and very basic definition of evolution : “a change in the characteristics of a group of items (caused by a change in the characteristics of at least some of the items so that the statistical distribution of the item characteristics in the group has changed).”

Related articles by Zemanta

• Virtual creatures and robots take on ‘a life of their own’

Image by Getty Images via DaylifeI did a search for free internet based collaboration tools. Result : the list is endless !!
I plan in the future to do a bit more structuring. But I think the list could already be useful to somebody.
Feel free to suggest anything that is missing and is free and internet-based.

Here they are in alphabetical order :

Adobe Acrobat

Amy Editor

Annotated Links

Assembla

backpack

Basecamp

bubbl.us

Buzzword

Collaber

CollabTRAK

Collanos

Comindwork

Cool Conference Live

CoSketch

dabbleboard

DeskAway

Diigo

dimdim

Dropbox

EtherPad

Flash Meeting

fmyi

Foldershare

Get Satisfaction

Google docs

laboratree

Linkstore

Mikogo

mind42

Mindomo

mindquarry

mindtouch

Nexo

Ning

onstageportal

ooVoo

project2Manage

ProjectPier

protonotes

Rypple

Scriblink

Show Document

SightSpeed

Skype

Spicebird

Spinscape

stixy

Task2Gather

Teamwork Project Manager Online

textflow

thinkature

TokBox

twiddla

Tynt

vyew

wetpaint

WIZIQ

writeboard

writewith

Yugma

Yuguu