What is AI for beginners

further education : Artificial intelligence for beginners

Artificial intelligence, or AI for short, is a catchphrase of our time. But it's more than that, because AI is already being used productively in companies in the form of machine learning (ML) and deep learning (DL) - we will discuss the difference below. In addition, a number of pioneering projects are to test the effect of AI in the company. For specialists and executives it means: Be there - or let the development pass you by.

AI has arrived in companies and careers

Even if the technical development is always left to the professionals who are allowed to adorn themselves with job titles such as data scientist or machine learning engineer, there are many tasks that have to be done in the context of these projects: around project management, test validation and Implementation of the prototype in productive systems.

Tagesspiegel Background Digitization & AI

Digital politics, regulation, artificial intelligence: The briefing on digitization & AI. For decision-makers and experts from business, politics, associations, science and NGOs.

Free test now!

Most start-ups, financial service providers and the established eCommerce companies have long since got AI up and running in their mobile applications, purchasing systems or the data processing of their online shops. These algorithms classify and forecast, suggest products or routes to customers, for example, or the ideal sales prices or budgets for social media campaigns to retailers. There will also be no more powerful fraud detection without AI. AI systems will take over most of the auditing in just a few years.

In industrial companies, AI systems are also used to improve order intervals in operational purchasing, to maintain machines predictively, to increase the energy efficiency of production logistics, to automatically monitor the quality of machine output - and generally to deal with anomalies in production or others Systems to recognize.

Specialists and managers with AI competence are needed

The applications mentioned still work today using rudimentary AI systems, some of which can be developed in-house and some can be purchased. In the future, these systems will be optimized, replaced by better ones or completely rethought. This requires good planning. In an AI system, the algorithm may be the more scientifically interesting part; In practice, however, it is primarily a matter of integrating the system into the IT infrastructure and the business process.

The behavior of the system must also be validated before, during and especially after the integration. This is not an easy task as we are dealing with extremely complex systems. To be able to carry out these activities confidently and to take responsibility for such projects requires general knowledge of the procedures and concepts of common AI systems.

Artificial intelligence, machine learning and deep learning

Buzzwords like Big Data or Industry 4.0 are often difficult to classify. However, AI, ML and DL must be clearly related.

Machine learning is a collection of mathematical methods of pattern recognition. For newcomers to the subject, it is useful to know that there are procedures that have their origins in stochastics and that work with frequencies of occurrence - i.e. probabilities - of certain events in the data and can create forecasts based on them. These are also easy to use when examining categorical data. Conditional probabilities and entropy are important concepts here.

And there are methods that come from algebra and weight similarities in a multidimensional vector space, compare them with one another or split them into subspaces. These algorithms are generally a little more powerful with a lot of numerical data, mostly iteratively optimizing and therefore tend to be more computationally intensive. Vectors, matrix calculations, differential calculus and optimization methods play an essential role in understanding these learning algorithms.

In both categories there are simple and also less easy to understand procedures. But all of these learning algorithms are able to solve many everyday and also very special problems. However, in a developer's practice, problems often arise when there is either too little data or too many dimensions of the data.

That is why the successful development of algorithms for machine learning processes is what is known as feature engineering. The attributes (basically the columns of a table) are selected as features from a set that appear to be the most promising for the learning process. For this preselection one uses again statistical procedures.

Deep learning (DL) is a sub-discipline of machine learning using artificial neural networks. While the ideas for classic machine learning were developed from a certain mathematical logic, there is a model from nature for artificial neural networks: biological neural networks.

Artificial neural networks learn connections

In artificial neural networks, an input vector (a series of dimensions) represents a first layer; this is expanded or reduced via further layers with so-called neurons and abstracted via weightings until an output layer is reached. This generates an output vector - a result key that stands for a range of values, for example, or for a specific class, e.g. B. Dog or cat in a picture. The neurons are like little lights that either light 100 percent or not at all or somehow in between. Through training, the weights between the neurons are adjusted in such a way that certain input patterns (for example photos of pets) always lead to a certain output pattern of glowing neurons (for example “The photo shows a cat” or “The photo shows you” Dog ”, with one output neuron representing the cat and the other representing the dog). The training ideally sets the weights in hundreds or thousands of iterations so that in the end the correct expenditure neuron always shines in a sufficiently high rate.

The advantage of artificial neural networks is the very deep abstraction of relationships. This takes place over several layers of the network, which can solve very specific problems. The superordinate name is derived from this: Deep Learning.

However, this also has a disadvantage: a trained neural network with more than two layers becomes so difficult to understand the way it works that it becomes a black box for us humans. Very complex tasks that are handled by artificial neural networks require the use of 20 or more layers with hundreds to thousands of neurons per layer - impossible to even begin to understand them.

For beginners, this has the advantage that you only have to deal with the theory and mathematics of deep learning to a very limited extent, because even experts are essentially limited to configuring the neural network from the outside and then testing it.

Deep learning is used when other machine learning processes reach their limits and also when separate feature engineering has to be dispensed with. The network itself learns which features play a decisive role in the problem-solving pattern to be trained.

Artificial intelligence (AI) is a scientific area that includes machine learning, but also knows other areas. Artificial intelligence not only has to learn, it also has to be able to store, classify and retrieve knowledge efficiently.

In practice, AI systems that (should) appear in robots or autonomous vehicles, for example, are hybrid systems that, in addition to the learning processes, are also equipped with fixed rules. These are to be regarded as the instincts of the system: Living beings not only learn from neurons, but also have innate instincts that cannot be broken - or only with great difficulty.

Can machine learning and deep learning be understood by ordinary people?

Classic machine learning methods are not particularly difficult to learn. The only really necessary step is to be able to get involved in mathematics or statistics to a certain extent. But here too - and now I'm going to break the myth - you don't have to be a math genius to understand at least the most well-known methods, and understanding here means understanding how they work in the background.

Learning deep learning is basically a little easier, at least if you can do without the detailed theory. Deep learning includes very complex algorithms that can be explained using complicated formulas. But you can get started with the practice, which is about improving the performance of the network by simply trying to add or delete layers.

This knowledge is sufficient to be able to understand the principles, strengths and limitations of AI. A data scientist or machine learning engineer, however, must go far deeper. An algorithm may be trained quickly and easily - but increasing its forecast accuracy from 94.3 to 95.6 percent can degenerate into real science and require a great deal of persistence.

Theory, programming and practice

To get started, the following applies: The classic methods of machine learning can best be learned in theory - not all, but most of the methods can be well understood on paper. No programming skills are required for this. It is important that the differentiation possibilities between monitored and unsupervised processes and between parameterized and non-parameterized models are understood. Naive Bayes and decision trees are suitable as an introduction to machine learning in combination with statistics. As an introduction to machine learning based on algebra, primarily the k-nearest neighbor algorithm and k-means.

If the general understanding is not enough and if you have some programming knowledge, you should familiarize yourself with the Python programming language and view programming examples with the free ML library Scikit-Learn. There are good tutorials on the Internet as well as good books in English and also in German. There are also good online courses, especially at the programming level.

There are also online courses, tutorials and books for getting started with deep learning. Here, however, it is recommended to only treat the theory of artificial neural networks superficially and to start practically straight away. There is no getting around the Python programming language and the TensorFlow and Keras libraries. TensorFlow is an (open source) library for deep learning published by Google, which is extended by the Keras library provided. In Keras, artificial neural networks can be created and executed with just a few lines of code.

This makes it easy to create multi-layer artificial neural networks and can be tested by everyone. Only major challenges are limited by the hardware, which is certainly limited for normal users.

There would be one more restriction: persistence and a tendency towards self-taught learning are certainly general requirements for anyone who cannot familiarize themselves with a month-long full-time course, but has to set it up on the side. Please don't give up - others have made it, tried it in practice and then wrote an article like this one.

The author teaches master's students in data science at the Berlin University of Applied Sciences. He is also Chief Data Scientist and one of the founders of DATANOMIQ, a consulting and service company for applied data science.

Now new: We give you 4 weeks of Tagesspiegel Plus! To home page