“Unless mankind redesigns itself by changing our DNA through altering our genetic makeup, computer-generated robots will take over our world.” - Stephen Hawking
I have always maintained that great analysis and insight is part art and part science. The science part is easy enough. One statistician uses much the same techniques as the next, and the software and tools available to them are all the same. Therefore, assuming you have acquired the correct skills and possess a mind able to understand mathematics and statistics, you can become a run-of-the-mill analyst.
The art is more difficult, less teachable. Firstly, it is all very well knowing how to analyse data, but what analysis will you do to it? In what direction do you look for insight? There is now so much data available, it cannot all be meaningful. So the skilful, artistic analyst looks for the meaning in the data and doesn’t pursue blind alleys. The output becomes useful.
And yet there is another layer still. The true artist among analysts can then apply this to improve the business of the client. This layer is interpreting what the data means in a commercial context. This is a true and rare skill.
I was therefore interested to read last month that scientists at Cornell University in New York have created software called Eureqa which is able to “do science”. In essence, it takes raw data collected from experiments, comes up with some likely explanations, tests them and looks for improvements. I know that computers have been doing the work of humans for many years using raw statistics to create information - in the US, for example, there is software which produces sports reports just from match statistics!
But this development looks very interesting indeed. It claims to use an almost Darwinian process when interpreting the statistics. Through a process of guessing some answers to the results, it then subjects these to a survival-of-the-fittest routine in order to arrive at the most likely explanation for the data. Of course, it is very early days for Eureqa and its ilk, but it does, nevertheless, give us food for thought.
Where will it take us in analysing marketing data? Inevitably, I think we have to assume that the science part I described earlier will become more the domain of the machine. But what is really interesting is that if the machine is able to test and learn as it goes, does some or all of the art also get automated? The machine can analyse the data quickly and follow many alternate routes, either rejecting a route or moving it forward for further analysis on a larger scale than the human.
If it can also produce the most likely explanations for the data at high speed, then perhaps my first level of artistic human intervention also becomes redundant. But I still feel that the high analysis art of commercial acumen will remain with the human. The commercial world itself is a product of the vagaries of the human character and an understanding of this is still essential to deliver the best advice from the data mined by the robot.
How long will that remain the case is anyone’s guess. Perhaps the sheer proliferation of data now available to companies means that automated solutions will be the preferred norm in the not-too-distant future. So, my fellow humans, we need to adapt and react to the rise of the robot. We need to ensure that what we can offer is the high-art of analysis, not run-of-the-mill charts and graphs.
I guarantee that, if we don’t, the robot will work longer hours for less money than you or I. Now that does make commercial sense.