Blog BI: Rise of the Machines - automated decisions Blog BI: Rise of the Machines - automated decisions

Blog BI: Rise of the Machines

In my previous blog Do it yourself! in this series of articles on Business Intelligence I opened the door to Artificial Intelligence, Machine Learning and Automated Decisions. Will The Terminator become a reality? That’s not how I feel. The fact is that we are going to hand over more and more decisions and entrust them to machines, both in the form of computers and robots.

With this blog I want to differentiate some of the fear that may appear and rightfully so.

Cruise Control

Years ago for the first time in my life I got into a car that had cruise control. A nice option, but one you have to get used to. This was especially the case in tunnels and in large bends, where one apparently has the tendency to unconsciously reduce speed a bit, where a cruise control does not. In a bend or tunnel I had the feeling that the car was accelerating, while the cruise control just kept the speed at a constant. Something to get used to.

My current car has an adaptive cruise control. Even better, especially in traffic jams, but also something to be getting used to. When a car moves in front of you, this system keeps a safe distance and adjusts the speed. At first my foot tended to go to the brake pedal. It took a while before I had confidence enough in this mechanism before I could suppress this tendency.

Campaign Obama

A few articles back I discussed Big Data. An example of applying this can be found in the first campaign, in which Barack Obama became president in 2009. The campaign team was assisted with an application of Big Data in the selection of houses, where it was useful to ring the doorbell. If a team member was in a certain street, the big-data analysis would have selected the homes in which people could still be influenced. After all it is of no use to call on people, who already were going to vote for Obama in the first place. But it would be useless with people, who absolutely were not going to vote for him. This allowed them to work very efficiently. It is worth mentioning that this analysis was done by a bunch of Dutch students.

BI and practice

A customer of us uses a certain dashboard in BI to gain insight into the exchange rate differences between the moment of closing and executing a contract. The financial employee checks these figures and then makes correction entries using these.
We have shaped this dashboard in the form of the application in JD Edwards, with which these lines can be copied and the journal entries are created. The next step is to automate this process and creates these journal entries via a web service in JD Edwards (called by Oracle BI). Schedule this process at night and a first form of Automated Decisions is a fact.

In my previous blog I referred to a video from Oracle. The example given here is about an insurance company. When a customer calls, the employee retrieves the data from that customer on his screen. An algorithm then shows a percentage of the probability that this customer would be interested in a new life insurance policy.

Initially, it will be repetitive series of actions based on historical data. But as soon as the intelligence and the predictive power of this type of algorithm become more and more reliable, then more and more automatic decisions can be considered.

Today’s BMW already contains software that uses the driving behavior, which combined with loads of data, logic and rules, to schedule an appointment at the workshop and orders the parts needed for that service.

Rise of the Machines

Will the horror stories of the Terminator film series become reality? Partially maybe. Robots will build robots. Software will develop software. Big Data appliances will predict. Increasingly advancing hardware, such as for facial recognition, will speed up and enable all of this.

The movie Minority Report is about the prediction of crimes which means someone can be arrested before this person has actually committed the crime. This is already possible with Big Data analysis. On the basis of patterns in a person’s life, the probability that someone will commit a murder can be calculated, including the when and the who.

A few years ago, a couple of American students had already devised an algorithm with which a person can be profiled based on the page Likes in that person’s Facebook. With 20 likes this algorithm can do this better than your colleagues, with 50 likes this exceeds your best friend and with 200 likes even your own partner.

It is a matter of time before we start talking about the decisions that machines make for us. Let me refer back to the cruise control in the car. It takes some getting used to …


Author:  Rick Brobbel
BI Consultant bij Cadran Consultancy