Categories
podcast-DG-en podcast-en

How computers manage to deal with language independently

Powerful language models have become available to the broad world public, making the success of computational linguistics evident: Computer architectures have been developed there that are capable of processing intuitive human language. The behaviour shown by these systems is hardly comprehensible in detail, but the results they deliver are all the more impressive: Assistance systems equipped with these language models can be controlled and generate output as if they actually mastered language themselves – and their potential applications go far beyond the sensational chat bots that dominate public debates. However, GPT and related systems have by no means emerged suddenly. They are the result of a persistent learning process: first attempts to map language into algorithms failed in the mid-twentieth century not only because of a lack of resources or because crucial machine learning methods had not yet been developed, but also because the theories on how to abstractly grasp and systematise the meaning level of human language were insufficient. So what do developers of modern systems do differently from the pioneers of computational linguistics?

Chris Biemann is Professor of Language Technology at the University of Hamburg, where he heads the Language Technology Group and the House of Computing and Data Science. In this episode of Digitalgespräch, the expert provides deep insights into the way modern language models are created and how they work, explaining linguistic theories that come into play. He makes comprehensible why the systems deliver such impressive results and describes what they can be used for in scientific contexts. With hosts Marlene Görger and Petra Gehring, Biemann discusses what resources go into the development of such systems, what happens when language models are trained on nearly the entire internet – and what tasks computational linguists face now that they seem to have achieved their great goal.

Episode 36 of Digitalgespräch, feat. Chris Biemann of Universität Hamburg, 16 May 2023
Further informationen:

Link to the textbook “Wissensrohstoff Text – eine Einführung in das Text Minig” by Chris Biemann, Gerhard Heyer and Uwe Quasthoff: https://link.springer.com/book/10.1007/978-3-658-35969-0
Link to the website of the House of Computing and Data Science: https://www.hcds.uni-hamburg.de/hcds.html

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Managing manifold data – the example of biodiversity research

Research on plant and animal biodiversity faces a technical challenge: data collections generated observing nature and in the environmental or life sciences are huge and constantly growing. However, data types and data sets differ greatly, depending on the research field and practical circumstances. For biodiversity research, however, heterogeneous knowledge must be consolidated and also shared – to solve ecological problems, one needs the large pictures. In concrete terms, the challenge can be described as: How to extract as much data as possible from countless spreadsheets, handwritten documentation, satellite images, living organisms or dried plants, so that researchers can access this resource conveniently via an online portal? One thing is clear: this is a mammoth task. And in addition to technical challenges, there are problems of data law and science policy to solve.

Dr Barbara Ebert is the managing director of the German Federation for Biological Data. She coordinates the development of a platform for research data on biodiversity research in the “NFDI4Biodiversity” project of the German National Research Data Infrastructure (NFDI). In this episode of Digitalgespräch, the biologist and data management expert explains the hurdles that have to be overcome and describes the sources from which data come together, what they are collected for and how they become usable for further research. With hosts Marlene Görger and Petra Gehring, Ebert discusses where negotiation processes take place, which needs are prioritised and whose interests must be taken into account.

Episode 35 of Digitalgespräch, feat. Barbara Ebert of the German Federation for Biological Data, 4 April 2023
Further informationen:

Link to the data management portal of the German Federation for Biological Data: https://www.gfbio.org/

Link to the website of the “NFDI4Biodiversity” project of the German National Research Data Infrastructure (NFDI): https://www.nfdi4biodiversity.org/de/

Link to a presentation on the compilation of data for the Red List of fish species: https://www.youtube.com/watch?v=AeO_PONc3oA

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Machine learning in environmental monitoring

Rapidly advancing climate change has consequences for ecosystems and landscapes. Observing and understanding these changes is crucial for minimizing damage through smart action. Environmental monitoring that provides reliable data on the condition of forests, soils and peatlands – with as few gaps as possible – requires cooperative research which combines knowledge about ecological concepts with modern, computer-aided procedures, providing methods and technologies to extend specific date obtained from certain areas to larger scales: Machine learning techniques can be used to correlate data that foresters, landscape managers, and nature lovers find visible and measurable in the field with satellite and drone imagery of larger spatial areas. In theory, spatial and temporal gaps in the data could be closed using those methods – provided the mathematical models behind them are understood and the initial data with which the “artificial intelligence” learns actually matches the ecological question.

Hanna Meyer is Professor of Remote Sensing and Spatial Modeling at the Institute of Landscape Ecology at the University of Münster. In this episode of Digitalgespräch, the expert explains how environmental computer scientists work and what typical tasks and questions are. She describes what data is needed to correlate satellite and drone imagery with real ecological systems and how machine learning helps to fill gaps in the data. With hosts Marlene Görger and Petra Gehring, Meyer discusses what the limits of this mathematical extension of field data are – and the dangers of trusting the models too blindly.

Episode 34 of Digitalgespräch, feat. Hanna Meyer of Universität Münster, 14 March 2023
Further informationen:

Link to the article “Qualität globaler Umweltkarten auf dem Prüfstand” in wissen|leben (WWU Münster): https://www.uni-muenster.de/news/view.php?cmdid=12772

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

What can small and large drones do? On the automation of airborne devices

Popular hobby or threatening combat device: the public perception of so-called “drones” seems to be shaped primarily by these two areas of application. As flying film and photo devices, they are operated mainly by amateur pilots for fun. However, in the lower air levels these inhabit, they are getting company: more and more drones are put to professional, civilian use. And the latter, sophisticated airborne machines differ not only in shape, size and weight, but also require new regulations and safety considerations. The civilian use of modern drones could be helpful for example in medicine, agriculture, environmental protection or for sea rescue. But in order to achieve this goal not only technical but above all regulatory and infrastructural challenges must be addressed. After all, all of these aircraft – including hobby drones – are part of civil air traffic.

Uwe Klingauf is Professor of Flight Systems and Automatic Control at the Technical University of Darmstadt and has been working in the field of aeronautical systems engineering for many years. In this episode of Digitalgespräch, the expert on automated flight systems talks about the state of the art, exciting current trends and challenges for research and development. He explains which safety considerations accompany the introduction of drones in urban areas, which processes have to take place in this and which applications are most promising. With hosts Marlene Görger and Petra Gehring, Klingauf discusses new use cases for automated aircraft, aspects that will determine which developments will prevail – and what has become of the hype surrounding parcel drones and flying taxis.

Episode 33 of Digitalgespräch, feat. Uwe Klingauf of Technische Universität Darmstadt, 21 February 2023
Further informationen:

Link to the information page “Drone Flight” of the German air traffic control Deutsche Flugsicherung: https://www.dfs.de/homepage/en/drone-flight/

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Digital credit scoring: How data analytics decide to whom to lend and not to lend money

When we open a bank account, apply for a loan or sign a mobile phone contract, we are usually asked to agree to a credit check – in Germany for example with Schufa. Those who request such a score about us hope to get a reliable statement about whether we are likely to meet our financial obligations. But how does this information come about? Big Data and Artificial Intelligence make it possible to use completely new strategies for this question, with which more individual, possibly more precise or even “fairer” scores could be created. This is where FinTechs come in, which – unlike Schufa – could basically use our entire digital footprint for their scores: Automated processes find patterns and correlations with which aspects of all areas of life can be translated into financial data. Whether the results, which include not only income and payment behaviour but also musical tastes and jogging routes, always correspond to reality is one question – another is whether personal characteristics play a role that are actually affected by the ban on discrimination. Because, as studies show time and again: AI models seem to systematically discriminate against socially disadvantaged groups. And loan sharks can also profit from this automated perpetuation of social injustice, depending on the legal situation.

Katja Langenbucher is a professor of civil law, commercial law and banking law at the House of Finance at Goethe University Frankfurt am Main. One of her research focuses is the usage of artificial intelligence in the financial sector. In this episode of “Digitalgespräch”, the law expert explains the considerations behind the development of new types of AI models for scoring, which problems they raise or might help to solve, and where regulatory needs arise. Together with hosts Marlene Görger and Petra Gehring, Langenbucher discusses differences between scoring providers, also in international comparison, and which justice problems require an open democratic debate, also at the EU level.

Episode 32 of Digitalgespräch, feat. Katja Langenbucher of Goethe-Universität Frankfurt am Main, 31 January 2023
Further informationen:

Link to Katja Langenbucher’s guest article „KI-basiert ermittelte Kreditausfallrisiken mit Vorsicht zu genießen“ in Börsen-Zeitung: https://www.boersen-zeitung.de/kapitalmarktforschung/ki-basiert-ermittelte-kreditausfallrisiken-mit-vorsicht-zu-geniessen-91ff697a-673e-11ed-a8ee-76a419d2158f

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Biochemistry meets computer science: How we can store digital data in DNA

Up to now, there are no forms of long-term digital archiving: the longest-lived hard disks and tapes are intact for no more than 50 years. After that, the data that was stored on them is lost. And even if chips and hard disks seem to be getting smaller and smaller, their compactness will reach natural limits at some point. At the same time, more and more important processes are taking place in the digital realm, we are collecting more and more digital data and developing new ideas and concepts to integrate information technologies into everyday life and technical processes. Our culture is also expressed in the digital, digital values and works are created. So the need for innovative storage media that can be used flexibly on the one hand and last for centuries and millennia on the other is there. One promising candidate in the search for solutions is DNA. How is it possible to translate digitality into this biochemical substance? And what new possibilities do then arise?

Prof. Dr Robert Grass researches and teaches at the Department of Chemistry and Applied Biosciences at ETH Zurich, where he works in particular on making DNA usable as a storage medium: He has co-developed a process in which DNA remains preservable in tiny glass beads for many millennia – and at the same time can be read when needed. In this episode of Digitalgespräch, the scientist and inventor explains how this is done, what challenges exist and what future visions he and his colleagues have for the development of this new storage technology. He describes possible applications for industry and business on the one hand and the archiving of digital as well as digitised cultural assets on the other. With hosts Marlene Görger and Petra Gehring, Grass discusses analogies between natural and technical processes and which difficult tasks absolutely must be solved if archives – especially digital ones! – are to be not only intact but also understandable in the distant future.

Episode 31 of Digitalgespräch, feat. Robert Grass of ETH Zurich, 20 Dezember 2022
Further informationen:

Link to film of the European Patent Office presenting and explaining the work of Robert Grass and Wendelin Stark as part of the European Inventor Award: https://www.youtube.com/watch?v=–4B0Pg4pf8 (English and French subtitles available)

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Looking into your own financial future: Digital Pension Transparency

Who has a feel for or let alone knows the financial situation they will face when old? Especially when retirement is still a few decades away, it seems impossible to really plan ahead for it. And so we wait. And again and again, we put off the question of whether our money will be enough. Especially when we fear there might be a significant gap. In fact, many people don’t realise that they have to become active themselves in order to make a living in their old age. And when one realises the problem of a “pension gap”, what can actually be done? The situation is utterly complex : savings, assets and possible insurances are just the start. Inflation and the development of the housing market also come into play. For non-experts, the topic easily comes across as too demanding – hence, a typical case for digital aids. Is it possible, factoring in data, constraints and individual factors, to compute forecasts and even alternative scenarios for someone’s pension?

Andreas Hackethal is professor of finance at the House of Finance at Goethe-Universität Frankfurt a.M. The economist researches household finance. And at Goethe-Universität, he is developing an elaborate app for simulating one’s own financial circumstances in old age – the “pension cockpit”. In this episode of Digitalgespräch, Hackethal explains why it is so difficult for many people to get a good idea of their own financial future and how an app can help to close such knowledge gaps. He describes how the complex app project came about and what is important when it is implemented. With hosts Marlene Görger and Petra Gehring, Hackethal discusses how it affects your decisions today when you know their effects on your future pension, what opportunities this opens up for individuals and families – and what political consequences it can have to make pension gaps transparent.

Episode 30 of Digitalgespräch, feat. Andreas Hackethal of Goethe-Universität Frankfurt a.M., 30 November 2022
Further informationen:

Link to the article on the “pension cockpit” in the magazine Brigitte (in German): https://www.brigitte.de/academy/finanzen/rente-berechnen–so-geht-es-12798828.html

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Gaming culture for everyone: people, debates and a billion-dollar market

What may once have been a gaming scene of insiders has long since been adopted by the full spectrum of society. Across all age groups and social classes, most Germans play video games – to relax or to shorten waiting time, as a serious hobby or even as a profession.  But not everyone who plays games would also like to describe themselves as a gamer. Although video games are recognised as a cultural asset, their public perception still seems to be dominated by the prejudice of the male, rather young gamer who spends far too much time with Shoot ’em ups – which in the worst case make people aggressive, but in any case are a waste of time. The large group of gamers and developers, however, is much more open and diverse; the gigantic range of video games is correspondingly versatile. So what do the gaming worlds look like today? What happens  apart from the infamous first-person shooters? And is gaming always “just” about entertainment?

Rae Grimm is Head of Digital Publishing at Webedia Gaming GmbH and responsible for their magazines GamePro, GameStar and MeinMMO. In this episode of Digitalgespräch, the expert for video game cultures describes which facets gaming offers today and what characterises different scenes. She explains what constitutes innovation, what is important to gamers and developers and what innovations and debates occupy the gaming world. With hosts Marlene Görger and Petra Gehring, Grimm discusses the social significance of video games and video game criticism, the gaming industry – and the uneasy feeling that chatting with AI-based, digital “friends” can give you.

Episode 29 of Digitalgespräch, feat. Rae Grimm of Webedia Gaming GmbH, 8 November 2022
Further informationen:

Link to Rae Grimm’s report “Wie ich ein Wochenende nur mit KIs geredet und fast den Verstand verloren habe” https://www.gamepro.de/artikel/ki-mental-health-verstand-verloren,3378382.html

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en

Digital playing fields in music production

It has been ages since listening to music did require being in the same place as the musicians themselves. However, digitalisation has absolutely transformed the production, availability and consumption of recorded music. And alongside with it, the music market: we have become used to as good as any piece of music being available at any time, and – at least in private – practically free of charge. If we wish to experience music as something exclusive and special, we have to seek or create such opportunities consciously and intentionally. With the possibility of recording and distributing music via streaming platforms, however, it is not only our relationship to music as a cultural event that has fundamentally changed: Digital technologies are bringing about a variety of upheavals in music production itself; a development that is perceived, evaluated and embraced – or refused – in very different ways by musicians and fans alike.

Sociologist David Waldecker researches and teaches at the Media Studies Department of Universität Siegen and has intensively studied how music is produced in audio recording studios. In this episode of Digitalgespräch, the expert explains the significance of the studio and its respective technical equipment for music production, both historically and in a contemporary context. He describes the possibilities that digitality offers – both for amateur music makers, who can now independently produce their own music, and for professional musicians and producers. With hosts Marlene Görger and Petra Gehring, Waldecker discusses the tensions that can arise between digitality and authenticity and how digital technologies shape our listening habits. And, after all, what importance does the distinction between “analogue” and “digital” actually have when we engage in making, listening to and experiencing music?

Episode 28 of Digitalgespräch, feat. David Waldecker of Universität Siegen, 18 October 2022
Further informationen:

Link to the article “Machine Learning in Context, or Learning from LANDR: Artificial Intelligence and the Platformization of Music Mastering” by Jonathan Sterne and Elena Razlogova:
https://journals.sagepub.com/doi/10.1177/2056305119847525

Link to David Waldecker’s book “Mit Adorno im Tonstudio” on the sociology of music production:
https://www.transcript-verlag.de/978-3-8376-5701-2/mit-adorno-im-tonstudio/

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.

Categories
podcast-DG-en podcast-en Uncategorized

Hacker Attacks and IT Management: Insurance against Cyber Risks

The number of attacks on IT systems has been increasing for years. They target small as well as big companies, state institutions as well as private individuals. The damage they cause can be enormous, not only in financial terms: in digital-ised environments, resilient IT is a key requirement for trustworthiness. Effective and reliable operation of production and supply chains depend on it and so does the data security of customer, patient and business partners. Everyone who consciously engages with digitality knows this and is aware of associated risks. Nevertheless – as studies show time and again – in many cases too little is in-vested in IT security. And although there are insurances specifically against “cyber risks” on offer by now, by no means do all companies whose integrity depends on IT security purchase them.

Florian Salm is an expert for cyber risks at Gothaer Allgemeine Versicherung AG and a lecturer at the University of Hamburg. Ulrich Greveler is Professor of Applied Computer Science at the Rhine-Waal University of Applied Sciences and, as an IT security expert, also a consultant and assessor for cyber risks. In this episode of Digitalgespräch, these two experts explain from a technical and insurance perspective what constitutes “well made” IT – and how companies can implement it. They explain how insurance companies help – not only in the event of an incident, but also in closing existing security gaps before something happens. With hosts Marlene Görger and Petra Gehring, they discuss why IT security is still difficult to implement in many companies, who bears responsibility for the security of systems in a networked society – and what it means that some risks of future technologies and large IT projects are not insurable.

Episode 27 of Digitalgespräch, feat. Florian Salm of Gothaer Allgemeine Versicherung AG and Ulrich Greveler of Rhine-Waal University of Applied Sciences, 27 September 2022
Further informationen:

Link to Ulrich Greveler’s website: https://www.ulrich-greveler.de/english
Link to survey “Gothaer KMU-Studie 2022: Cyberangriffe größte Bedrohung für Mittelständler”: https://presse.gothaer.de/pressreleases/gothaer-kmu-studie-2022-cyberangriffe-groesste-bedrohung-fuer-mittelstaendler-3182062

all episodes of Digitalgespräch»

Creative Commons Lizenzvertrag

The podcast is in German. At the moment there is no English version or transcript available.