Most read articles
|1||Language: An important component of inclusion|
|2||Career and family – that’s (not) possible!|
|3||The role of technology in the business model of law firms|
Do you belong to those managers ready to invest in AI-projects during the next one to three years? Or do you prefer postponing this decision by three to five years? According to a new survey, 93% of executives from high-growth enterprises agree that AI investments should be considered urgent. In low growth companies, 64% of managers postpone this decision. Especially in low-investment settings, where competition is rather small, one likes to emphasize that, as competitors are inactive, the urgency of being active itself decreases. This is particularly dangerous when it comes to AI, because technology scales enormously. If a competitor – such as a new start-up, you’ve never heard of, or Amazon – has suddenly captured a considerable market share, it is difficult or even impossible to win it back. (This is particularly true in the service sector and slightly less in the industry). That is why you don’t want to fall behind on this issue; and if you do, this has to be a conscious choice rather than an action taken out of convenience.
Executive School Programmes:
Furthermore, studies proof that on management levels, superficial technical terms often characterize the discussion of artificial intelligence, while a profound understanding of the various algorithms and necessary data is usually missing. Even though this fact is rather disappointing, it is generally regarded as sufficient, since expensive specialists can always be considered to resolve further questions. However, we are convinced that this assumption can be an obstacle to the successful implementation of AI-based solutions. As a manager, you need to know more about AI. This does not mean that you should be familiar to programming, but the ability to grasp basic concepts of AI technology needs to be existent.
What is artificial intelligence (AI) anyway? What does machine learning (ML) mean?
Algorithms of machine learning predict target values on an individual level. This prediction can be for single customer as well as for machines. These target values are crucial to commercial success and efficient process design. Take a bank for instance that wants to predict whether someone is credible. The databases is filled with information about customers who have either paid their loans without any problems, or had to be depreciated. All defaulting borrowers will then be marked 1 and all others will be marked 0. This variable is the target variable, usually called Y variable. Data on other variables such as place of residence, gender, age, education and income are simply called X variables. These X variables are important, as they influence the Y variable. An algorithm learns from the data that results of a combination of X-variables and repayment results (Y) and calculates a repayment probability. Since the algorithm learns these patterns independently from the data, this particular process is called machine learning.
Another application example would be to predict whether a machine would fail and/or require maintenance. In this case, the X-variables are temperature, running time, product type etc. and the algorithm can predict when the machine needs maintenance before it really stops. Other examples of X and Y variables are:
- Pixels of images as X and the state of emotion of a pictured person as a target variable; or
- Digital audio signals as X- and identification of a customer’s request on the telephone as target variable.
As already mentioned algorithms of machine learning and of AI are self-learning. Algorithms are “trained” before they are used and learning takes place while training. At the very beginning, an algorithm is “stupid”, meaning one hundred percent unaware of facts and circumstances. Thus, he makes a lot of mistakes. During training, these errors serve as feedback and the algorithm adjusts its behavior according to the errors previously made. In this way, the algorithm detects patterns in the data and stores them in a statistical formula. This formula codifies the algorithm`s “knowledge”, which is the added value that it provides. Different algorithms distinguish when it comes to successful recognition of certain patterns.
You may have wondered about the difference between machine learning and AI. Experts can argue about this for ages. There are several opinions among scientists on this distinction. It is, however, clear that machine learning can be considered a subdivision of AI. This particular subsection has been proving humongous success for the last ten years. Latest trends such as face recognition, autonomous driving or the prediction of the likelihood of loan repayment are all applications of machine learning.
It is obvious, that an algorithm learns from data. This is the reason why data is often referred to as the raw material and new oil of our economy. However, this supports a wrong assumption. Data doesn’t “flow” in companies and can’t easily be tapped in order to achieve high profit, as it would be the case for oil. A more precise metaphor can be derived from the field of agriculture. Imagine you want to grow a specific agricultural product. The choice of seeds you grow obviously depends on how the final product is to be used. It goes without saying that you have to plough and fertilize the soil. If you ignore this and you don’t take care of your plants, you will have a poor harvest. The same goes for the data in your systems. It does not automatically have a high potential of value creation. You first have to think about which data you need in order to successfully use AI, and you have to “cultivate” the systems through the creation of suitable data systems.
The follow up article to this blog will cover why managers should understand basic concepts of AI. There are several decisions that shouldn`t be left to data specialists only.
Do you prefer watching a video than reading this blog? We then recommend our Webinar.
picture from unsplash