With examples from our daily lives, Paul Hiemstra made the abstract topic of big data come alive for an almost full house at the Leeuwarder Courant in Leeuwardern. Netflix, fake news, YouTube and Google ads all came up. ‘These techniques all work on the basis of big data.’ But big data alone does nothing and has no value, according to Hiemstra. ,,Data only becomes interesting when you assign characteristics to it. For example, the movies on Netflix, they have characteristics such as fiction, action, romance. If you link these characteristics to other characteristics that say something about the viewer, such as the time at which the film is seen and which other films he watches, then you can make predictions. These predictions Netflix dishes out to the viewer as suggestions for watching the next movie or series.”
cats
To be able to make predictions, a computer has to learn all kinds of things. ”Just like a child learns,” Hiemstra explains. His 1-year-old nephew learns because his mom and dad point things out to him and name what it is. ,,That way he learns everything about cats and eventually recognizes a cat. We do the same thing with computers. We show the computer pictures of all kinds of cats until he understands the characteristics of a cat. Then, when it sees a picture of a horse, the computer infallibly knows how to say: in all probability this is not a cat.‘ Hiemstra explains that ’showing‘ is just as important for a child as it is for a computer to learn. The computer stores everything and after some time ’knows' how to unerringly distinguish between all kinds of animals. Hiemstra calls such a computer model an algorithm. With it, characteristics attached to data suddenly become valuable.
Cow
That is the field of work of Anne Gerben Terpstra. He shows how his company can measure the quality of cow's milk using various colors of light and thus make predictions about the cow's health. “By linking all kinds of characteristics to certain colors of light with smart sensors, we can measure the fat content in milk and hundreds of other things. Previously, a chemical analysis took three days and then you received a result on a maximum of 5 measured points. With our method you have a result of hundreds of measuring points within 6 seconds.”
Privacy
Big data can therefore help us further and make our lives easier. Terpstra: "With our measuring method, the University of Amsterdam has shown that you can predict with 96 percent certainty whether someone has cancer from human blood samples." This kind of information is nice to know for a good and quick treatment, but it also has a downside. Terpstra agrees. "Of course you don't want health insurers to find out this kind of information about you, because that can mean that you are excluded from insurance or that you may have to pay a significantly higher premium." According to him, this method is already being used in car insurance, where sensors are used to record people's driving behaviour. If you drive recklessly, you have to pay a higher premium. “They want to increase road safety in this way.”
The use of big data offers not only a ‘dark side,’ according to Hiemstra, but also great opportunities. ’’The technology is far ahead of the public debate and legislation. That's why I want to share my knowledge and make people aware of what's going on so they can form an opinion.''
This and the other 4 lectures can be viewed on a special webpage of the Leeuwarder Courant / Future Makers.