Data mining and machine learning

"AI decisions follow pre-determined tracks"

10. May 2024 by Theresa Dirtl
Why artificial intelligence is actually not that intelligent, why most data remain unused and how to keep track of all the rapid developments. Data expert Claudia Plant provides her insights in the Rudolphina interview.
Data expert Claudia Plant considers AI neither artificial nor very intelligent. In this interview, she explains why and provides an insight into her research in the fields of medicine and renewable energies. © Pexels

Rudolphina: Claudia Plant, you are Head of the Data Mining and Machine Learning research group at the University of Vienna. What exactly is data mining?

Claudia Plant: Data mining is a sub field of artificial intelligence and aims to automatically extract knowledge from very large amounts of data. We are collecting huge amounts of data in almost all areas of life today, for example, in medicine or in business. In fact, the data we actually use is only the tip of the iceberg, since the majority of the data collected simply remains unused. My research group explores how we can make accessible personal data, in particular – for example, in the field of medicine – and use them for the benefit of humans to improve treatment options, for instance.

Rudolphina: Many people have reservations with regard to the use of, especially, personal data. With good reason?

Claudia Plant: Many people fear that they could become transparent beings so to say. With regard to our use of the Internet, this fear is not entirely unjustified: If we want to use apps and services, we first have to agree to the terms of use including all the fine-print, mostly without knowing exactly what we are currently signing up for. With our consent, we practically grant the companies the right to apply certain algorithms to our data.

Rudolphina: Have companies intensified this use of data along with the rapid developments in the field of AI?

Claudia Plant: Well, social media have certainly accelerated research in the field of AI. Since we generate and provide a large amount of data every day, it is interesting for companies to tap into these data to fulfil their business goals. We produce such huge amounts of data that we are no longer able to analyse them manually or by using simple statistical methods. To prepare large amounts of data in a way that people are able to understand and use them, we need to make use of algorithms, for example, from the fields of machine learning and data mining.

Save the Date: Panel discussion on the semester question, 17 June

Data expert Claudia Plant will also be a guest at the final event on the semester question "Do we know what AI will know?", which will take place on Monday, 17 June 2024, at 6 p.m. in the Main Ceremonial Hall of the University of Vienna.
Science and technology researcher Helga Nowotny (author of "Die KI sei mit euch") will give a keynote speech and then take part in a panel discussion with:

  • Claudia Plant, Professor of Data Mining and Machine Learning, University of Vienna
  • Nikolaus Forgó, Professor of Technology and Intellectual Property Law, University of Vienna
  • Beatrice Blümel, lawyer and alumna of the University of Vienna
  • Markus Tretzmüller, founder of the start-up Cortecs and alumnus of the University of Vienna
You may also read
Rudolphina Experts: AI Policies
The regulation of artificial intelligence comes with both challenges and opportunities at the interface between technological progress and social values. A perspective by legal scholar Iris Eisenberger.

Rudolphina: The term ‘artificial intelligence’ by itself creates a feeling of unease in many people...

Claudia Plant: I believe that the term ‘artificial intelligence’ is indeed very unfitting, since the system is neither artificial nor is it particularly intelligent. It has been created by humans and did not develop from itself. AI is based on algorithms that collect data, for example, from the direct environment in case of self-driving cars, and make decisions based on these data. But these decisions are practically pre-determined, since the algorithms have been invented and programmed by humans.

Overall, AI is of course a very complex system, but the individual components follow laws that have been determined by humans. Our current challenge is that deep neural networks consist of comprehensible components but are overall highly complex due to the large number of different components. We are currently exploring how we can understand them again. But I want to emphasise that AI systems are based on algorithms. And all algorithms are based on comprehensible components from mathematics and computer science.

I believe that the term 'artificial intelligence' is indeed very unfitting, since the system is neither artificial nor is it particularly intelligent.
Claudia Plant

Rudolphina: You and your team are focusing on applications that could directly benefit human beings. Can you give us some examples?

Claudia Plant: My research group mainly develops data mining procedures that could be used in medicine or for renewable energies. In medicine, for example, we can find individualised treatment options for patients. For this purpose, it is necessary to collect a wide range of biological as well as medical data. Renewable energies are an entirely different area of interest. In this field, we attempt to predict how much wind or solar energy we can generate in a certain period. This could help improve the network capacity and would thus enable us to better coordinate the customers’ requirements and the amount of renewable energy produced.

In future, it will be possible in many areas for AI to take decisions that are so accurate that they almost seem like human decisions.
Claudia Plant

AI promises faster help in case of depression

Claudia Plant and her team are working on a method to detect early on whether patients suffering from depression respond to medication. The fact that more than 30 per cent of those affected do not respond to the common medication is one challenge in this context. It often takes weeks or even months of trial and error until patients receive suitable medication. Claudia Plant’s team has developed an innovative method to tackle this issue: They use algorithms to analyse data from the patients’ EEGs. After only one week of treatment, these algorithms are able to predict whether the medication will actually be effective or not. This enables a faster change of medication. Claudia Plant explains, "Our algorithms are able to detect hints in the EEG which doctors are not able to detect with the naked eye."

Rudolphina: These are two of an incredibly wide range of AI applications. In your view, which sectors or fields could currently benefit most from the developments in the field of data mining and machine learning?

Claudia Plant: We see enormous potential in the life sciences, for example, in the search for new medicinal substances. Ever more accurate measuring instruments enable the life sciences to generate huge amounts of data, which we can search and analyse by means of data mining.

Also in the business sector, of course, plenty of new business models based on the combination of the Internet and artificial intelligence are developing. Unfortunately, Austria and the German-speaking region in general are not playing a key role in this regard. The major Internet corporations are based in the US and Asia. Therefore, I personally find it important to promote our local start-up scene, because we need to develop AI business models based on European values. Otherwise, I see the risk that a very large share of development funds are distributed to few stakeholders, i.e. the major tech companies.

You may also read
The future with AI
Canadian computer scientist and Turing Award winner Yoshua Bengio is one of the world's foremost experts in AI and deep learning. On 7 May he spoke at the University of Vienna about potential catastrophic scenarios from advanced AI systems and how to prevent them. We had the opportunity to ask him a few questions beforehand.

Rudolphina: Algorithms and AI are developing extremely rapidly. How do you keep up with the state of the art as a researcher?

Claudia Plant: I am aware that I only know a small proportion of it all. Therefore, I specialise in certain niches, in which I am actually able to compete internationally in research. At the same time, I only have limited resources and thus investigate issues for which I do not necessarily need to work for a multi-million tech company.

So how do I keep up then? Mainly thanks to the young members on my team, who are practically at the forefront of developing and programming these things. I provide assistance in the process and learn a lot myself in doing so.

Using algorithms to understand the power consumption of Austrians

In a current project on renewable energies, Yllka Velaj and Claudia Plant are the first to investigate data of around 250,000 households in Austria collected by smart meters. Smart meters are intelligent, electronic devices that measure energy consumption every 15 minutes and transfer the information to the grid operator. Such data have been available worldwide for 15 years, but have only been investigated in small-scale studies using classical methods to date. This project aims to develop new approaches to identify the actual energy consumption at certain times (to see when and where are peaks).

Rudolphina: Do you consider it a problem that the major tech companies are able to invest much more money than science and research?

Claudia Plant: Yes, especially research on deep neural networks requires enormous amounts of data and computing resources. In comparison to the large tech companies, universities have only limited capacities for GPU clusters. These are computers which are equipped with graphic cards and are able to do a wide range of calculations simultaneously. In addition, tech companies do not always publish their algorithms and data since they are part of their trade secret. It is always alarming when research is not open. The more people are able to actively contribute to research on AI, the better we can ensure that the progress we are making will benefit all of us. We require more diversity in AI research. It is essential that people with different cultural and social backgrounds and gender identities jointly develop AI algorithms.

The major internet corporations are based in the US and Asia. Therefore, it is important to promote our local start-up scene to develop AI business models based on European values.
Claudia Plant

Rudolphina: Do you have tips for students and interested readers how they can stay up to date in the field of AI?

Claudia Plant: Our University is actually very open to IT topics. This is also evident from the current semester question, which I find wonderful. There are many lectures on the topic. For example, we offered the lecture series "Machines that understand" last semester and the Department of Contemporary History offers a lecture series on the challenges and opportunities with regard to artificial intelligence this semester. These lecture series are open to students of all degree programmes, allowing them to gain a fundamental understanding of the issue and enabling them to later apply this knowledge to societal discussions.

In my opinion, every student should at least know what artificial intelligence is, what algorithms are and which principles they are based on. Of course, I would be happy about one or the other student who has a deeper interest in the issue and embarks on a career in computer science.

Anniversary year at the Faculty of Computer Science

The degree programme in Computer Science at the University of Vienna celebrates its 50th anniversary today and the Faculty of Computer Science turns 20 this year. On the occasion of this anniversary, computer scientist and Turing Award laureate Yoshua Bengio, one of the most highly cited AI researchers worldwide, hold a lecture on 7 May.

  • Further information about the 2024 anniversary of the Faculty of Computer Science
  • Further information about the degree programme in Computer Science at the University of Vienna
  • The "Studieren Probieren" (have a taste of studying) event provides the opportunity to take a glimpse of the courses of the degree programme: on 22 May and 5 June 2024

Rudolphina: Finally, I would like to ask you our current semester question: Do we know what AI will know?

Claudia Plant: I say, we do. We can know it, even though it will be difficult since the entire field is so incredibly wide and diverse. But it is all about algorithms, and we are able to understand algorithms . In future, we will see plenty of development cycles and it will be possible in many areas for AI to take decisions that are so accurate that they almost seem like human decisions. In the process, we always need to remain cautious and use AI carefully to the benefit of humans.

Rudolphina: Thank you very much for the interview.

© Barbara Mair
© Barbara Mair
Claudia Plant is Professor of Data Mining and Head of the Data Mining and Machine Learning research group at the Faculty of Computer Science. Her research interests include parameter-free data mining based on information theory, integrative data mining methods for heterogeneous data sets and application-oriented data mining in biomedicine, neuroscience and environmental sciences.