The digitalization is the number one topic in today’s society, be it private or business environment. One of the focus areas is Artificial Intelligence (AI) – a highly dynamic field that is constantly evolving, supported not only by large companies such as Telekom, but also by the global Open Source Community with its Open Source Software (OSS).
There is an inflationary number and variety of different definitions around AI. These range from “A machine behaves like a human being” to “A machine thinks like a human being”. And “A machine thinks rationally.” up to “A machine behaves rationally”. It can be clearly seen that there are two different forms – the weak AI, which deals with individual abilities that a human being can perform, and the strong AI, which describes a machine that can do everything a human being can do. The strong AI is not yet realistically tangible and exists only on a theoretical level. Thus, the AI includes methods that deal with enabling human behavior of a machine.
To achieve this, Machine Learning (ML) plays an important role, which is a branch of AI and allows machines to learn in an efficient way through algorithms and data themselves. There are three analysis methods – descriptive, predictive and prescriptive – and three different models – Supervised Learning, Unsupervised Learning and Reinforcement Learning.
Since AI and ML are often used interchangeably, depending on the background of the respective person, another subject area, a sub-area of ML, must be introduced – Deep Learning (DL), a neural network to efficiently process large data sets on several levels. The three terms – AI, ML and DL – are hierarchically and logically interrelated.
Picture 1: Logical classification AI, ML and DL, based on
(Goodfellow, et al., 2016) (Chui, et al., 2018)
Deep learning techniques have achieved publicly breakthroughs in recent years, for example in speech recognition, the recognition of objects and their states in images or video material (cancer recognition!) and many more.
However, the true potential of the algorithms in the field of AI is not yet fully exploited. On the one hand, this is because of the sheer size of that topic and on the other hand, does more advanced techniques within Deep Learning more hardware power. By using the Open Source Model the availability of such technologies can be increased and downsides decreased, so companies can be provided with more efficient and more stable solutions.
Open source software is a model that allows people of any background to view, copy, modify and redistribute the source code of this software based on a specific open source license. The already mentioned global open source community makes its solutions available to the public on platforms such as GitHub and thus creates a collaborative basis to work together and time-efficiently on innovative solutions. Because it is not about constantly reinventing the wheel, but about keeping up with the times and remaining competitive. Exactly this point confirms that Open Source offers many opportunities for companies of all sizes to use.
The Telekom Data Intelligence Hub offers the most important open source solutions on one platform in addition to the data marketplace with its connected AI workbench, provides easy access via the Jupyter Notebook and R Studio environments and benefits from the innovative development of the open source community to always provide the most important state-of-the-art OSS.
During an expert survey, the following OSS were selected and ranked as the most important and most efficient solutions in the field of AI.
Picture 2: Ranking of Open Source Software for Artificial Intelligence (Wilhelm, 2018)
Exactly these solutions are easily provided by the Data Intelligence Hub, without the need to implement the different OSS individually – all on one platform, fast to use and always the latest version to stay innovative and competitive as a company.