![](https://www.lll.tum.de/wp-content/uploads/2023/04/ai-data-science-machine-learning-visual.jpg)
Artificial Intelligence, Data Science and Machine Learning – what lies behind the buzzwords?
When the conversation turns to the latest developments in information technology, the terms Artificial Intelligence (AI), Data Science and Machine Learning are bound to pop up. What these terms actually mean and in what context they should be used is however often unclear. Even though these “hyped” terms are often mixed together, they can be clearly distinguished from each other. It is important to differentiate the concepts behind AI, Data Science and Machine Learning in order to facilitate general understanding of these complex topics and to reduce false expectations, hopes and fears.
This is why we asked three of our experts from the field to comment on the connections, misconceptions and possible new skills required when dealing with these technological developments.
Dominik Bösl is CTO & Managing Director of Micropsi Industries, which develops software for industrial robots in dynamic environments. He is also Professor for Digital Sciences, Automation and Leadership at Hochschule der Bayerischen Wirtschaft (HDBW). He holds a diploma in Computer Science from the University of Augsburg and an MBA degree from the University of Pittsburgh. He is a regular lecturer at different universities and author of technical and scientific publications. At TUM School of Education, he is researching „Technology, Robotic & AI Governance”: the ethical, moral, socio-cultural, -political and -economic implications of technologies, as robotics, automation and artificial intelligence on humankind. He has founded among other initiatives the Robotic & A.I. Governance Foundation in order to foster the interdisciplinary discourse about these topics.
Jakub Cichor is a research associate at the Chair of Research and Science Management, Head of Focus Area “Educational Technologies in Leadership and Expert Development” at the TUM Center for Educational Technologies (TUM EdTech Center), and a member of the TUM Neurophysiological Leadership Lab (TUM NeLeLab). He completed his B.Sc. and M.Sc. in Informatics: Games Engineering at the Technical University of Munich with an emphasis on serious games, gamification, and game-based learning. His research focuses on educational technologies at the intersection between human-computer interaction and organizational behavior, specifically on investigating social robots and AI in leadership positions and implementing Virtual Reality applications for leadership development.
Thomas Mair is research associate at the Institute for Machine Tools and Industrial Management, TUM School of Engineering and Design. He holds a B.Sc. and M.Sc. degree in Mechanical Engineering. His master thesis addressed the use of Artificial Intelligence for predicting internal irregularities in friction stir welding. His research focuses on additively manufactured damping concepts for use in vibration-prone components. For this purpose, the design of the structures and the topology optimization of the components are investigated by using artificial intelligence. He currently supervises the lecture “Artificial Intelligence in Production Engineering” at TUM.
What is the relationship between AI, Data Science and Machine Learning?
Dominik Bösl: “It all starts with data. Without it, the processes and technologies mentioned are either useless or not applicable. Data Science describes all processes for collecting, storing, homogenizing, processing, validating and – depending on the definition – analyzing data in the context of data analytics. In Machine Learning, algorithms interpret the data. They recognize patterns and correlations from which solutions to problems can be derived. Artificial Intelligence (AI) attempts to go one step further: Through appropriately elaborate algorithms, processes and technologies – such as neural networks and transformer systems – the aim is to achieve the most ‘intelligent’ and autonomous behavior possible. Computer systems, such as in autonomous vehicles, assistance systems for decision-making or (service) robots, should be able to react autonomously to their environment and make decisions. However, there is not yet a generally valid term for intelligence – which leads to many discussions about what should or should not really be considered artificially ‘intelligent’.”
Jakub Cichor: “Data Science, AI and Machine Learning are often considered together to analyze and interpret large amounts of data in an effective and efficient way. Data Science is fundamentally concerned with data and regularly uses methods from AI and Machine Learning to analyze data. AI is used to implement intelligent systems, with Data Science being just one of many application areas of AI. Machine Learning, on the other hand, is a subfield of AI that focuses on methods that learn from data and then use the acquired information to interpret data.”
Thomas Mair: “The three terms are closely related. Artificial intelligence (AI) is the generic term for intelligent, self-thinking machines and programs. Machine Learning (ML) is a part of AI and includes models or algorithms that learn from data to make decisions and predictions. Data Scientists deal with the extraction and analysis of data, they use AI/ML to generate knowledge.”
What are the greatest misconceptions about AI, Data Science and Machine Learning?
Jakub Cichor: “A major misconception is that the results of these technologies are objective and unbiased. In machine learning in particular, one problem is that humans generate most of the training data. This leads in turn to a data basis influenced by human tendencies and biases that get adopted by the system. A well-known example is an algorithm in the medical field that learned to primarily look for rulers on images for cancer diagnosis instead of actually identifying tumors. The reason for this error was quite simple: rulers are generally placed on medical images if a tumor is present in order to more accurately represent its size.”
Dominik Bösl: “One of the biggest misconceptions is that AI – currently especially the complex language models around ChatGPT – are a cure-all for any unsolved problems in computer science, robotics, automation, image recognition, text generation, for smart computer assistants, and more. The progress shown in recent weeks and months in the use of extremely large-scale language models and transformer-based neural networks is impressive and certainly groundbreaking for some application fields. Automatically generating – or simply proofreading – everyday text in emails, business letters or for text summarization will save a lot of time. However, we cannot – and should not – expect, hope for, or dream of creative input, the development of meaningful, content-related innovations, or even automated research. This is precisely where one of the biggest risks of AI lies: the dream of an intelligent digital colleague who can map all the mundane activities of everyday life, understand us and react according to our preferences is too good to be true. In the medium term, however, these expectations are unrealistic – at least for the time being. Each of the hyped technologies has its intended use and place of application; none, though, is universally applicable. For example, ChatGPT can understand language and derive the likely expected actions from it – but it cannot automatically calculate a robot’s movement plan and maneuver the gripper arm to the appropriate place in space to grab the desired apple.”
Thomas Mair: “AI will replace many professions. Of course, artificial intelligence will have an impact on the structure of the working world. Nevertheless, this statement cannot be generalized. Let’s look at the example of a software developer. The chatbot ChatGPT is able to generate entire code structures based on the developer’s specifications within a very short time. This does not mean that the software developers are replaced. Their activities shift from structuring and implementing the code to formulating the task for the chatbot. Other activities, such as planning and debugging the code, remain unchanged. This means that jobs are upgraded and not replaced.”
What kind of skills should you have to work with AI, Data Science and Machine Learning?
Thomas Mair: “The capabilities required depend on the application layer. Three examples are shown below:
1. No special skill is required for everyday use of ready-made AI tools. Developers provide an intuitive interface, enabling ease of use.
2. For technical use cases, there should be an advanced understanding of the problem and possible solutions in order to be able to verify the results.
3. When implementing AI for new use cases, a data scientist needs sound programming skills and a basic understanding of data analysis methods to be able to adapt and train the models.”
Jakub Cichor: “It is important to have knowledge about and a basic understanding of the technologies to be able to interpret the outputs in a meaningful way while being aware of the pitfalls of these systems. Responsible use of the results can open up many new possibilities, while uncritical adoption of wrong results can cause many problems.”
Dominik Bösl: “When dealing with new, disruptive technologies the most important skills certainly include fundamental openness and willingness to deal with the subject. This does not necessarily mean (in relation to AI) being able to mathematically grasp and calculate the underlying models yourself – but the differentiation of concepts mentioned at the beginning helps in understanding and assessing opportunities, risks and applicability in a personal context – be it in private, business or a scientific environment. Personally, I would like to reverse the information loop: all stakeholders, whether from science or industry, who develop relevant technologies, should be obliged to explain their results in a generally understandable way. Only if we enable and promote meaningful, informed discourse across society in the first place, will we succeed in promoting participatory innovation.”
We thank our experts for their deep insights into the topic and for their efforts to explain these complex terms. There is more behind AI, Data Science and Machine Learning then it might appear at first glance.
Are you interested in learning more about disruptive technologies like Artificial Intelligence, Data Science and Machine Learning? Join our masterclasses on several related topics or learn new skills with our certificate programs.