Why should one learn to analyze data

AI, machine learning, data analysis - the pillars of future technology

Tom Becker

Jupiter, Juno and Minerva were the leading trio in ancient Rome and were of the utmost importance alongside one another and with one another. A similar picture can currently be drawn in the modern data age, but no longer of images of gods, but of the newly created supporting pillars of future technology: data analysis carried out by humans, artificial intelligence and machine learning.

EnlargeThe amount of data grows and grows - how should we deal with it?

Companies and institutions need the combined strength of the trio in order to master the growing data boom together. For example, if we wanted to download all the data from the Internet, we would need 181 million years according to current calculations - and the amount of data is growing exponentially. Back in 2013, IBM said that 90 percent of the data had been created in the previous two years.

What are the characteristics of artificial intelligence, machine learning and data analysis? And why should companies use the potential of these three technologies in combination?

Artificial intelligence

Machine learning (ML) and artificial intelligence (AI) are often used interchangeably in public debate. However, there is one crucial difference: ML deals with the automation of processes and the "learning algorithm".

AI, in turn, mimics human decision-making structures and tries to make (independent) decisions based on data.

However, the desire for human-like intelligence in robots or computers did not only exist since the computer age. As early as ancient Greek mythology, Talos is said to have automatically thrown stones at ships to defend Crete as a bronze giant. And human-like "automatons" can be found in the literature long before the first computers, for example in E.T.A. Hoffmann's "The Sandman".

Today we divide AI into two categories: applied AI and general AI. Applied AI focuses on one aspect. The less widespread general AI going towards it claims to be able to cope with theoretically different tasks. And it is precisely in this area that machine learning has developed.

Machine learning

ML describes a system that is taught to learn from experience. It consists of three main components:

1. Model: System that makes predictions or identifications.

2. Parameters: signals or factors that the model uses to make decisions.

3. Learning system: system that adapts the parameters - and thus the model - by considering differences in the predictions compared to the actual result.

With these three mechanisms working together, ML is able to perform reliable and consistent analyzes and to learn from examples and to generalize them. In addition, the learning system even identifies and stores patterns, regularities and laws during the learning process. As a result, it is then able to transfer the experience afterwards, i.e. also to assess previously unknown data. There is a difference between systems with symbolic approaches, in which knowledge is explicitly represented, and systems with non-symbolic approaches, in which knowledge is implicitly represented. The currently much discussed neural networks, for example, do not allow any insight into the solutions that have been learned.

Human-driven data science

The basic principle of data analysis is actually quite simple and consists in recognizing patterns from empirical values ​​and abstracting them from them - a deeply human characteristic. For example, a grandmother's ability to use her own life experience to give relationship advice for the decisions of others - a striking example used by the former Google data scientist Seth Stephens-Davidowitz in his New York Times bestseller “Everybody Lies”.

With the help of powerful computers and highly developed software, it is now possible to process and analyze amounts of data that would be inaccessible to us without technical aids. Until some time ago, special programming skills were required to analyze data volumes, but now even non-specialist employees have the option of analyzing data independently using special self-service software; they are so-called Citizen Data Scientists.

Programming-free platforms not only enable Citizen Data Scientists to analyze data, but also contribute to the fact that data analysis can run like a red thread through all areas of the company - and does not have to be limited to IT.

This is a critical factor in business success because, according to a survey by Forbes Insight and EY, companies with data analytics embedded in their overall business strategy are seeing growing revenues and higher margins.

Interplay of human-borne data analysis, ML and AI - a powerful triad?

Although some of the areas of AI and ML are currently still a long way off for data analysis, both are already being used in companies and will become even more important in the future - as confirmed not least by the Fraunhofer Institute in a study. AI and ML can support analysts especially with repetitive and stupid tasks. These activities, which are often very labor-intensive and not very demanding, have cost companies a lot of time and money.

AI and ML enable tedious manual tasks to be summarized in a few seconds and carried out automatically. The capacities freed up as a result can be used much more effectively, for example for the analysis of large amounts of data or careful investigations. By the way, the golden rule of thumb is: the more specialized the tasks, the more AI can be useful.

The precise interaction of human-made data analysis, AI and ML is important for successful results. Because as promising as the last two technologies sound, they have a major disadvantage compared to us humans: the lack of context. While automated analysis can be extremely effective, it is empty of meaning and is of no use if not understood. People still have to be able to interpret, interpret and read the results obtained. Professor Dr. Martin Ruskowski from the German Research Center for Artificial Intelligence was therefore carried away recently at a conference with the concise description: "Neural networks are stupid and can never replace 3.5 billion years of evolution".

That is why people with the ability to think analytically and critically come into play in meaningful and future-oriented data analysis. People think associatively and abstractly and can put results in the right context and interpret findings appropriately. Interestingly enough, the Citizen Data Scientists are sometimes even better suited than the actual Data Scientists, as they bring with them the valuable context knowledge from the respective specialist departments.

Companies are therefore well advised not to look at AI, ML and human-supported data analysis in isolation, but rather to connect, use and combine them. This triad will be able to work even more effectively and develop unimagined possibilities.

Just as Jupiter, Minerva and Juno together formed the highest trio of gods, this trio will also offer new possibilities for future technology and become the mainstay of future companies.