Artificial intelligence glossary: AI terms you should know

More and more managers, including those without a technical background, are coming into contact with artificial intelligence (AI) because of automation and digitalization. As a result, terms related to AI are often used too broadly or incorrectly. In the glossary below, you will find important terms related to artificial intelligence.

Summary

Artificial intelligence (AI) in practice

Artificial intelligence is already used in many areas of everyday life:

• Search engines make it easier to deal with the flood of information on the Internet.
• Analysis and forecasting of share prices are sometimes supported by artificial neural networks.
• We encounter voice assistants and facial recognition every day.

Artificial intelligence has been gaining importance for years and enables companies to operate more profitably and gain a competitive advantage.

DeepMind

DeepMind is a company that specializes in programming artificial intelligence. The company’s official goal is to understand intelligence. Unlike many other AI companies, DeepMind therefore does not have a predefined goal and is more flexible in how it applies AI to different problems.

AlphaGo and AlphaZero

DeepMind became known worldwide for AlphaGo, a computer program that plays only the board game Go.

Although all rules and information in Go are known to all players at all times and there is no random element such as dice, the complexity of Go is many times higher than that of chess. It is therefore impossible to calculate all possible move combinations. Go requires a certain amount of (artificial) intelligence.

In October 2015, AlphaGo defeated Fan Hui, a multiple-time European champion. That made it the first program to beat a professional Go player several times in a row under tournament conditions.

In 2017, an improved version of the program called AlphaGo Zero was released, which beat the old program 100 to 0. AlphaGo Zero developed its strategies autonomously. What is interesting here is that AlphaGo Zero’s learning phase took only three days, whereas the older AlphaGo version required months of training.

This shows not only how “intelligent” computer programs can be, but also how quickly research and development in the field of artificial intelligence is progressing. Processes and methods that we consider state of the art today will already be obsolete in five years.

It is essential for companies to stay alert, learn about new technologies and processes, and use them in a way that creates value. Industries such as automotive and finance have already used AI very successfully in recent years and will continue to do so. Artificial intelligence is no longer a technology reserved for large corporations. AI is also extremely important for SMEs if they want to remain competitive in the international market.

Artificial Intelligence (AI)

AI stands for the English term “Artificial Intelligence”. It is a subfield of computer science that deals with automating intelligent behavior and machine learning. In practice, it often tries to replicate certain human decision structures and, in some cases, improve them by using large amounts of data. The goal is to enable a computer to solve problems on its own. Today, AI can be found in every industry, for example in medicine, banking, or retail.

Machine Learning (ML)

Machine learning enables IT systems to recognize patterns and rules based on existing data and algorithms. In a sense, it generates “artificial knowledge” from experience. The insights gained from the data can be generalized using statistical methods and used for new problem solving or for analyzing previously unknown data. The decisive factor is the preparation and structuring of the data, because machine learning requires a clearly defined target.

Deep Learning

Deep learning, also called multilayered or deep learning, is a special method of information processing. Its functioning is inspired in many ways by how the human brain learns. Based on existing information and a neural network, the system can repeatedly combine what it has learned with new content and thus learn again.

Deep learning teaches machines to learn. As a result, the machine is able to make predictions or decisions. Decisions are confirmed or updated in a new run. In most cases, humans no longer intervene in the actual learning process.

The main difference between machine learning and deep learning is that deep learning can convert unstructured information such as text, images, and speech into numerical values. These extracted pieces of information are then used for pattern recognition or further learning. Classic machine learning, for example based on decision-tree methods, is not able to process this unstructured data in a meaningful way.

Smart Factory

A smart factory is at the center of so-called Industry 4.0. In the ideal case, it is a production and manufacturing environment that organizes itself without human intervention. Humans no longer need to intervene in the actual production process.

The basis of the smart factory is so-called cyber-physical systems and the intelligent networking of machines and products. The product itself communicates the information needed for manufacturing to the smart factory. Based on this information, the individual production steps are controlled until the desired final result is reached. In many cases, communication between products and equipment takes place wirelessly. The communication basis is the Internet of Things (IoT).

Artificial neural networks (ANNs)

Artificial neural networks form the basis of AI, essentially its infrastructure. ANNs are inspired by the human brain and can be used for machine learning and artificial intelligence. Such networks consist of an abstract model of interconnected neurons, whose special arrangement and links make it possible to solve application problems from fields such as statistics, engineering, or economics using computers.

ANNs are ideal for applications where there is little systematic solution knowledge available and a large amount of partly imprecise input information has to be processed into a concrete result. Typical applications are speech recognition and image recognition. When the structure is multilayered and hierarchical, we speak of deep learning.

Robotics

Robotics deals with the design, construction, operation, and use of robots as well as computer systems for their control, sensory feedback, and information processing. One key area of application is Industry 4.0, where industrial robots are used among other things.

Collaborative robots, or cobots (short for “collaborative robot”), are becoming increasingly important. Traditional industrial robots are being replaced or supplemented in more and more areas of industry by collaborative robots. Cobots are used together with humans in the production process and are no longer separated from employees by protective devices, as typical industrial robots are. Compared with traditional industrial robots, collaborative robots are more compact, more flexible, and easier to program.

In the private sphere, so-called service robots are used. Service robots are machines that provide services directly for people. A distinction is made between use by private individuals and in professional environments. In private households, vacuum-cleaning robots and lawn-mowing robots are already common.

Graphics processing unit (GPU)

GPUs, or graphics processors, consist of a large number of processing units (shaders) and rely heavily on parallel computing because 3D calculations can be carried out quickly and efficiently that way. This makes GPUs the preferred hardware basis for artificial neural networks. The term GPU was first used extensively by Nvidia to market the Nvidia GeForce 256 series, which was released in 1999. To this day, the company is the market leader in GPUs and the world’s most valuable chip designer.

Central processing unit (CPU)

A CPU, usually called the main processor or central processing unit, is optimized for the fast execution of sequential tasks. Parallel computing is, in practical terms, the exact opposite. CPUs existed long before GPUs.

Together, GPUs and CPUs form the basis for AI. Their importance becomes obvious in the current chip shortage, which affects thousands of companies worldwide and costs companies billions. Semiconductor manufacturers expect bottlenecks to continue until next year, 2022. Customers simply order more than the factories can produce.

Data Engineer, Data Scientist and Data Analyst

Although the terms are similar and the responsibilities can overlap, there are some differences: data engineers are involved in creating and maintaining complete data architectures, data scientists handle their complex analysis, organization, and interpretation, while data analysts usually focus on finding patterns in numerical data and using them to help companies make better decisions. You can find more information on this topic in our article “Data Scientists: New, innovative thinkers”.

Conclusion

There are countless terms related to AI. It is certainly not necessary to know all of them at once, but it is worth building up knowledge in this area step by step. AI is no longer a technology reserved for large corporations. It is also highly important for SMEs to remain competitive in the international market, and it is a field that will continue to accompany us.

Do you have any further questions, comments, or would you like us to provide more information on AI? Feel free to contact us anytime and we will be happy to answer your questions.