FIFA 17 in Real Life: How Nerds Are Taking Over Football

5/5 (3)

When I play FIFA I always take a look at the numerous statistics about my players. Wouldn’t it be great if football managers in real life have these statistics and functionalities? Football is one of the most conservative sports, which is strange since it is the most popular sport in the world. A Dutch company called SciSports is trying to change this as they are introducing new technologies to football clubs.

SciSports is founded by Giels Brouwer, an alumni of the University of Twente. One of their most successful systems is BallJames. BallJames can be described as an MRI-scanner which scans the field during a match. It consists of fourteen cameras which observe all twenty-two players (and the ball, of course). A computer is analysing all kind of data, for example speed of headers, shots (on goal) and distance covered. BallJames is making use of ‘voxels’ which are 3D-pixels. A football match can generate up to 60 terabytes of data, but that’s just one part of the story. The next step is to come up with recommendations which can help managers. The data analysts of SciSports can help with this, the bigger clubs like Manchester City have their own data team, but smaller clubs don’t have these resources.

SciSports is growing rapidly, they raised 1.35 million euros which they are going to invest in BallJames. Clubs which are using BallJames are Heracles Almelo, PSV and Vitesse. There is an increasing number of ‘laptop-coaches’, a famous example is Pep Guardiola which is extensively using data to improve trainings and the performance of players. But BallJames is not only useful for football clubs, it can also be used at sports like rugby or even crowd-control on airports. I think that every professional club needs such a system in the future, because it can give them some interesting insights.


Please rate this

AI and Us

5/5 (1)

In our day to day live we experience that our smartphones have some artificial intelligence (AI) embedded in the form of a personal assistant. This assistant, be it Siri or Google Now, can perform tasks that include looking up and then presenting information verbally. Other features include being able to dictate a text message or creating a calendar entry.

Sci-Fi movies gave us the idea of being able to talk to an AI like we would talk to a real person. The AI’s in the movies, however, have the processing power and memory of a server farm (or even quantum computers) at their disposal.

To create an AI that we regard as intelligent, we have to consider what we consider intelligent behavior. Since we want to create an AI that matches our intellect, we should look at the most intelligent species we know, humans. Humans are particularly good at recognizing patterns. We can train to recognize certain shapes faster, e.g. in mathematics or even art. Computers, however, must be taught to categorize patterns according to what we teach them.
Teaching the computer these patterns to imitate the capabilities of the human brain, is called deep learning and thus creating an AI.

Now that we know what the goal is, what does that mean for businesses? Is AI important to be a part of the digital mastery?

Certainly, companies like Facebook or Google are working on this technology with remarkable results in image and speech recognition.

Other markets are also following the trend of having an intelligent bot at your side, for tasks that seem too complicated or intensive for us.

One sector that follows mathematical rules and where the lifeblood is what a computer knows to work with are financial systems and the tremendous amount of data they process daily. So far they are mainly used to serve customers, much like a sophisticated chatbot (e.g. SEB in Sweden, Royal Bank of Scotland) that answers the questions of customers. Paypal uses the technology to categorize types of fraud, while in Korea an AI delivered a 2 percent return on invested funds. In the automotive industry, we are starting to have very sophisticated autopilots. Assembly lines are more and more staffed with robots that are faster and more reliable than human staff. The possibilities to implement a powerful AI seem endless.

But what about the other side of the coin? How far do we go when we are not limited by processing power or other resources anymore? Why do leaders in their field such as Elon Musk, Bill Gates, and even Stephen Hawking warn about AI? What does that mean for the concept of “business”?

Recommended readings:

Stephen Hawking, Elon Musk, and Bill Gates Warn About Artificial Intelligence

The fear index by Robert Harris

Please rate this

Flying Cars, when can we expect them?

5/5 (1)

A combination of an airplane and a car is a combination where people have dreamed of for hundreds of years. Jules Verne (1828 – 1905), a pioneer of the science fiction  genre, was the first one thinking of such a combination. However, he was a science fiction writer, a poet. How serious could he have been? Let’s move on to 1940, when Henry Ford stated the following: “Mark my word: A combination airplane and motorcar is coming. You may smile,but it will come” . Now, 66 years later, this ‘ridiculous’ statement is almost becoming reality.

AeroMobil, a Slovakian company which has been into the ‘flying car business’ for almost 9 years, has announced the AreoMobil 3.0 will be available in 2017. The Aeromobil 3.0 is the latest physical prototype of the integrated flying and road vehicle unveiled and tested in 2014 and 2015. There have been already successful flights as can be seen in the following video.

This is the flying car of the future. It’s not the Harry Potter kind of car I expected, but still, it’s a car that can fly. But are flying cars a serious replacement for the regular ways of transport? I’m skeptical, but it’s a begin. AeroMobil predicts that in 2030 flying cars can be available for regular customers. Of course, flying cars won’t be as cheap as a regular car, but much cheaper than you would expect. Vaculik, Co-Founder and CEO of AeroMobil, said he expects that the vehicle will initially appeal to luxury car fans and flight enthusiasts, and will have a price tag that also falls somewhere in between the latest offerings from Tesla Motors and a small plane of a “couple hundred thousand dollars.”

Besides the technological difficulty and the price, the safety of such a car has to be sufficient. What kind of drivers’ license would you need? Would you need airbags or should you wear a parachute? In The Netherlands, an average of 4.500.000 calls for road assistance is made. With a regular car, this isn’t a really big problem. It will just take some time and money and you’ll be on your way again. However, with flying cars? Mechanical failure? Pretty lethal.

So when can we expect to get one of these cars ourselves? I think the prediction of 2030 is pretty accurate concerning the technological and financial part, but for the safety part it’s too early. There are so many possible safety issues to be kept in mind. However, technology can move faster than you would expect.

What do you guys think? Can we drive/fly ourselves a flying car in 2030?




Sources and other interesting material:



Please rate this

Deep learning: the new edge of machine learning

5/5 (5)

Machine learning has reached a new frontier in technology: deep learning, which is an algorithm inspired by how the human brain works. It is a new way to predict the future and makes machine learning closer to Artificial Intelligence. From a technical perspective, a lot could be discussed to understand the architecture and engineering of deep learning. However, this article aims to describe some applications which rely on this new algorithm.

This new technology arose few years ago, and many tech companies have already made big investments in it to improve some of their products, as self-driving cars. Deep learning is mainly about making better representations and creating models from large-scale unlabeled data: in simple terms, it relies on creating patterns of objects and classifying them based on common features. So far, the most relevant application that deploys deep learning techniques is image recognition.


Image recognition in our daily activities

As of 2012, deep learning allowed computers to recognize traffic signs twice as good as humans. Google made several experiments with it, and was able to map every single location in France in two hours. Facial recognition is also based on this approach: when Facebook is suggesting to tag specific people in the photos that you post, deep learning algorithms are in action. This is completely different from what happens with Google image searches: it is not anymore about grouping pictures based on the text that describe them on a website. It is about actually recognizing the objects in the photo, and grouping photos based on the features characterizing them. Experiments have shown that, when people are asked which caption they prefer, those created by computers win over those by humans.


Image recognition in medicine

Deep learning is having the most disruptive impact in medical imaging. For instance, a company called Enlitic examines millions of images to automatically learn to identify a disease. They are able to generate a new diagnostic test in 15 minutes, and get 90% of classifications right. Recent research not only has made predictions better, but also could create new insights on science. A study from Stanford looked at cancer tissues under magnification and discovered that cells around the cancer are as important as those characterizing the cancer itself: this is the opposite of what pathologists have been told. Many similar discoveries are being done by people who do not even have a medical background, even though the results are potentially changing the world of health. Plenty of new collaborations between medical companies and data scientists will keep on rising, and they may help to build a better world.



TED talk for Enlitic:


Facial recognition:

Deep learning:

Deep learning:

Please rate this

Here is why I do not believe in the Hype cycle and you should not either

5/5 (1)




When one talks about technology progress and which technologies are more promising and will be popular, there will be somebody mentioning the Garnter hype cycle and how tech #x will be the future. Let me tell you: I do not believe any of that. I do not even get why anybody created that framework or cycle at all. On the Gartner website (1), they say that this hype cycle should be used for businesses looking to launch a new product, so that they can balance risks with commercial success.  Is it really so? Does the framework work? Let’s take a closer look at it:

  1. Technology and commercial success have little in common. Technologies can be available yet there is a high chance nobody is willing to pay for it. Maybe the technology is not useful, maybe it is not integrated in the right product, or maybe just the product is not marketed well. In any case, the consumers might not buy the technology. But wait. In the Gartner hype cycle, there is nothing said about consumers, market potential or any whatsoever commercial metric. An example that came to mind is videocalling. I remember when I was 12 years old and this Italian provider started producing phones with this camera with the main assumption that people would enjoy videocalling each other. The company invested much effort and money in developing a product that could do support that (Apple and Smartphones were not there yet), creating a network and promoting 24/7 the new technology. It never took off. Now, 11 years later people do not videocall that much either. Or let’s take a more practical example. Recommendation systems or predictive analytics are based on alrgorithms that were mostly developed before the 80’s. It is true that computers back then were bigger and slower, yet consumers data were already there. Supermarkets have had fidelity programs for decades. So why is the technology booming now? The technology is not, its applications are.
  2. There are many inconsistencies, also in the jargon. If you look at the hype cycle in 2010, predictive analytics almost reached plateau of productivity. In 2016, Machine Learning is just over the peak. I wonder how is this possible ? Predictive analytics is based on machine learning, so either Gartner had a really different definitions of those two concepts or they made a mistake in calculating the cycle. To hide that, they just use some different buzz word and voila, everything looks right again. Further more artificial intelligence is much behind machine learning, although machine learning represents pretty much 90% of what artificial intelligence is (2). Does that make sense? Not to me. Moreover, many of terms they use mean are really cryptic. I guess that those can be really handy for business executives giving speeches or for some new young startup founds trying to get investors on this new technology that nobody really knows.
  3. Progress is more like a tree than a cycle. Well, if you think of a cycle you would not come up with anything looking like the Garnter cycle but most likely with a circular form. Apart from that, the only circular element in it is this flow of expectations increasing and decreasing. Yet comparing the Hype cycle from year to year I see that many technologies appear and disappear. There is not really a story or a continuity to check how they develop over 3 or 4 years. This aspect defies the concept of the cycle and lets us wander where all these technologies ended up in? They did not go through the cycle stages or they just died? We will never know.

To me it seems like the Gartner’s analysts are a bit confused. As in they do not have standard metrics to base their hype cycle on (3). I could imagine that some companies finance them to put some buzz words on exactly that point of the curve, to launch new startups or this amazing service which will boost their revenue streams. I do not see any added value in this Gartner’s cycle or any measurable basis to even call it a model. Do you?



1 –



Please rate this

What went wrong (In the 2016 election)?

5/5 (1)

Some of you already gathered themselves together after the shocking news which was spread on the 8th of November, but some of us (including myself) are still are trying to wake from this bad dream full of racist pronunciations. As all of you know on the morning of the 8th of November in the Netherlands the news was announced that Trump will be the 45th president of the United States. The traditional prediction models were far off this year which were applied throughout the last decade. The forecasting of FiveThirdyEight which used three different models predicted Hilary Clinton with a 71.4% change of winning against 28.6% for Donald trump, getting most swinging states wrong. A similar result was shown by a lot of traditional projection polls.

Forecasting in 16th
while some people are still discussing what went wrong the some possible solution can be found on the internet. The traditional ways of predicting the presidency no longer applies in 2016 with the social media having such a big impact on our daily lives. When looking at the prediction based on non-traditional variables such as:  the money spent on Social media adds and   media attention.  Compared to Donald Trump, Hilary spent 5 fold what Trump spent in June. This more traditional way of measuring predicted that the amount raised and spend for their customer engagement would present a similar positive effect on the customer engagement. This was at the end definitely not the case.

What was actual the best way of predicting the 2016 election was using the social media attention with a 1:1 correlation. The way to winning for Donald J. Trump was to look outside the traditional news publications towards mostly posting on social media website and applications. Compared to the year before that Donald Trump has a chocking 78% of his retweets which were from the public compared to Clinton (0%), Bernie Sanders (2%) and Obama (3%).

Anger as projection of his Victory
Another explanation behind the misprojection  was that the media did not fully cover or understand the anger which was presented by some of the American people. After the elections and already during the elections it became clear that a lot of people were upset by the selective recovery and the betrayal in the trade deals. They saw it as a threat towards their jobs and also as disrespect towards themselves . The wrong polls of the presidential election can also be explained as fuel to the fire which was already burning. That the polls motivated people which would not otherwise have voted to take action against the predictions which was that Hilary would win presidency.

Will this year’s elections influence all the presidential elections to come and should the shift be made to different ways of predicting the presidential outcome because of the digital shift of most nations?  Or will this be a one-time fluke compared to previous elections and  all elections to come?

But the end we are all hoping that Donald J. Trump will positively surprise us in the year 2017 during his reign as 45th president of the United Sates.

Please rate this