Sustainability is not an optional feature. Targeting economic, social, and ecological issues through sustainable development has become a must, not least since the publication of the UN’s Sustainable Development Goals in 2015.
Computer Science (CS), and more prominently and recently, Artificial Intelligence (AI), is hailed as the solution to many sustainability problems (and yes, it is!). Still, energy efficiency and CO2 reduction must be considered to ensure fair and ecological use of resources. Hence, dealing with informatics and sustainability involves two faces of the same coin: Sustainability with CS, often referred to as “Computational Sustainability”, and Sustainability for CS, commonly known as “Sustainable CS”.
This lecture series is the first of its kind in Austria, with all Austrian CS departments and faculties participating, under the auspices of Informatik Austria. We want to raise awareness for the utmost importance of sustainability in CS, demonstrating research solutions to various problems, and triggering developments for a sustainable future.
The lecture series will be held in the winter semester 2023/2024, every Monday from 17:00-18:00. All lectures will be offered online and will be recorded (see below). We also have one hybrid lecture on January 8, 2024, on-site at TU Wien and online.
Whether you’re a student, a professional looking to expand your knowledge, or a citizen interested in sustainability issues – this lecture series is open to anyone. Switch on your laptop and join us!
Students from participating universities can register to get credits for the series. Please refer to your respective institution for registration.
TU Wien students can register for VU 194.155 “Sustainability in Computer Science” via TISS. Depending on your study program, the lecture series is a free elective or a “Wahlpflichtfach”, please refer to TISS for further information.
By Daniel Gruss, TU Graz on Oct 9, 2023. Global ICT electricity consumption is already beyond 11 percent of the worldwide electricity production and still increasing. By 2030 it may reach around 25 percent. Previous approaches to improve efficiency and performance have often sacrificed security, leading to disastrous security issues like Meltdown and Spectre. Patching just these two vulnerabilities increases power consumption on affected computers by a seemingly harmless 5 percent. By 2030, this may be more than 1 percent of the global electricity production by just a single out of thousands of patches. This development is not sustainable, and in this talk, we will discuss both the problem and potential revolutionary solutions.
By Andreas Uhl, PLUS on Oct 16, 2023. With the increasing importance of sustainable solutions in ICT, social media platforms have been increasingly criticized for their significant carbon footprints. We will have a look at corresponding results and will analyze the methodology how these estimations have been obtained, partially revealing that the perspective from which such investigations are being conducted is significantly biasing the results.
By Nysret Musliu, TU Wien Informatics on Oct 23, 2023. Real-life problems in domains as diverse as health care, supply chain, production and education are frequently very challenging and their solutions impact people involved as well as the resource consumption and efficiency of operations. Optimization and AI provide robust techniques that can be employed to effectively address such problems in a variety of sustainable applications. This lecture gives an overview of various AI and optimization techniques, including problem-solving and machine-learning techniques. Through case studies in scheduling and planning, as well as food waste reduction, we will demonstrate how these techniques can be applied to solve complex problems in sustainable applications.
By Philipp Gschwandtner, Universität Innsbruck on Oct 30, 2023. As the demand for high-performance computing continues to surge, the environmental impact of powering and cooling these massive computing infrastructures has become a pressing concern. This talk attempts to unravel the critical intersection of computational power and sustainability within the HPC domain, focusing on energy-efficient hardware, multi-objective optimization and other relevant hardware and software practices.
By Schahram Dustdar, TU Wien Informatics on Nov 6, 2023. Modern distributed systems deal with many uncertain scenarios, where environments, infrastructures, and applications are widely diverse. In the scope of IoT-Edge-Fog-Cloud computing, sustainability issues for designing energy-efficient distributed systems arise. Leveraging neuro-inspired principles and mechanisms could aid in building more flexible solutions able to generalize over different environments. A captivating set of hypotheses from the field of neuroscience suggests that human and animal brain mechanisms result from few powerful principles. In this talk, we discuss some fundamental research activities with the goal to make progress in the design and execution of sustainable digital ecosystems.
By Krzysztof Pietrzak, ISTA on Nov 13, 2023. The Bitcoin blockchain achieves consensus in an open setting, i.e., where everyone can participate. This was believed to be impossible, the key idea to make this possible was to use computing power rather than some kind of identities for voting through “proofs of work”. Unfortunately, this approach is not sustainable: the Bitcoin blockchain burns roughly as much electricity as a country like Austria. We will outline sustainable alternatives for achieving a Bitcoin-like blockchain, with a focus on using disk-space instead of computation and how this is realized in the Chia Network blockchain.
By Ivona Brandic, TU Wien Informatics on Nov 20, 2023. In the first part of this talk, we will discuss the concepts of sustainable IT. We will present novel approaches for sustainable fault tolerance and trustworthy geographically distributed AI applications utilized on systems ranging from the size of a smartphone to a warehouse-scale data center. Furthermore, we will discuss the novel concept of hybrid classic/quantum systems as a response to the increasing demand for computational resources. Scientific applications are nowadays utilizing different types of hardware accelerators including GPUs, TPUs and NPUs. With the arrival of the post-Moore Era and the rise of quantum computing, we are experiencing new opportunities but also challenges when trying to integrate quantum computers into the well-known computational continuum. This integration into so-called hybrid systems promises resource efficiency at a new order of magnitude. We will discuss challenges regarding the data encoding, noise, transpilation and the high heterogeneity of quantum architectures which necessitate the development and integration of additional hardware/software layers. In the second part of the talk, we will present several use cases, where IT can be utilized to combat climate change and help us to develop a more sustainable world including sustainable watershed management through IoT-driven artificial intelligence and satellite-based monitoring of livestock in the Alpine region.
By Moritz Grosse-Wentrup, Universität Wien on Nov 27, 2023. AI systems and machine learning models are increasingly deployed in high-stakes environments, e.g., in medical decision support or loan approval, where ethical and legal considerations require models to be interpretable. Recent research in interpretable machine learning (IML) and explainable AI (xAI) has made great strides in providing methods that reveal the inner workings of machine learning models. Several key challenges, however, remain to be addressed. In particular, these challenges include distinguishing between interpreting the model and using the model to interpret the data-generating process, and providing intuitive explanations suitable for domain experts rather than for AI researchers. In my presentation, I will provide an overview over IML/xAI methods and show how we can resolve (aspects of) these challenges by adopting a causal perspective on IML/XAI, which interprets models in the context of the data-generating process.
Electric Load Disaggregation as a Means for Increasing Energy Awareness and Reducing Energy Consumption
By Wilfried Elmenreich, Universität Klagenfurt on Dec 11, 2023. When the EU introduced smart meters, the idea was to give consumers more information about their power consumption, thus raising energy awareness. Tailored appliance-level feedback is expected to lead to more careful use of energy and thus to savings of around 12%. Despite the potential benefits, implementing a network of small measurement devices, each per appliance/socket, has raised concerns due to the added energy consumption and intrusive visibility of smart adapters impacting user experience. Non-Intrusive Load Monitoring employs computer algorithms to extract appliance consumption data from a single, smart meter measuring the entire system’s consumption. While this concept has been around since 1985, when George W. Hart (MIT) proposed it in a report for the Electric Power Research Institute, recent advancements in computer science have propelled its potential. This talk explores the significance of load disaggregation and its relevance in today’s context, where smart meters and small embedded devices like the Raspberry Pi offer sufficient computational resources. The presentation delves into the evolution of load disaggregation algorithms, from traditional bin-packing techniques and Hidden Markov Models to the cutting-edge application of deep neural networks powered by machine learning. As these techniques continue to evolve, the talk also addresses the pressing need to address privacy issues and develop comprehensive, long-term interfaces that engage consumers in a captivating manner to monitor their power consumption effectively. In conclusion, the talk highlights the potential of Electric Load Disaggregation in increasing energy awareness and reducing energy consumption. Leveraging advancements in computer science, this technology promises to substantially impact sustainable energy practices, empowering consumers to conserve energy and promote a greener future.
By René Mayrhofer, JKU Linz on Dec 18, 2023. Digital information and communication services offer the potential for significant energy savings: video conferencing instead of trans-continental flight, file transfer and streaming instead of shipping physical media, or cryptocurrency instead of transporting gold and paper money. But what is the real energy balance of these digital options? In this talk, we try to quantify such comparisons based on estimates of the energy consumption of various digital services.
By Dieter Kranzlmüller, Leibniz Rechenzentrum München on Jan 8, 2024. The demand for computing is steadily growing, and the surge of artificial intelligence has further increased the need for computing capacities. This is in particular visible in supercomputing centers worldwide, where more computing leads to more electrical power consumption. As a consequence, energy efficiency is not just a nice-to-have option, it is a must-have in order to deliver more performance for simulation and modeling, artificial intelligence and machine learning. The Leibniz Supercomputing Centre (LRZ) is at the forefront of energy efficiency, providing its leadership class systems with direct hot-water cooling. In fact, the holistic approach of LRZ addresses all aspects of energy efficiency, from the point where electrical energy enters the building to the usage of waste heat, and all the future technologies involved in providing energy-to-science.
By Kaspar Lebloch, Universität Wien on Jan 15, 2024. Today’s smart devices are tomorrow’s paperweights. Particularly Internet of Things (IoT) devices have gained a reputation for being unreliable in their overdependence on mandatory services and their providers. Some consumers actively avoid smart home devices or leave them disconnected due to privacy or reliability concerns. In this lecture, we explore the untapped sustainability potentials in retrofitting existing legacy hardware with open, local IoT technology in different ways and try to envision a path toward a more modular, personalizable, and upgradable IoT.
By Franz Wotawa, TU Graz on Jan 22, 2023. In this talk, we are going to discuss the intersection of software engineering research and sustainability. We start with discussing existing approaches in software engineering for improving sustainability and in particular reducing the energy consumption of applications. Afterward, we outline the application of sustainability to software.
PDF / 3.86 MB / 2023-10-09-Slides-Daniel-Gruss.pdf
PDF / 2.28 MB / 2023-10-16-Slides-Andreas-Uhl.pdf
PDF / 2.5 MB / 2023-10-23-Slides-Nysret-Musliu.pdf
PDF / 3.53 MB / 2023-10-30-Slides-Philipp-Gschwandtner.pdf
PDF / 3.38 MB / 2023-11-06-Slides-Schahram-Dustdar.pdf
PDF / 5.06 MB / 2023-11-13-Slides-Krzysztof-Pietrzak-MOD.pdf
PDF / 6.14 MB / 2023-11-20-Ivona-Brandic.pdf
PDF / 4.63 MB / 2023-11-27-Slides-Moritz-Grosse-Wentrup.pdf
PDF / 2 MB / 2023-12-11-Slides-Wilfried-Elmenreich.pdf
PDF / 2.66 MB / 2023-12-19-Slides-Rene-Mayrhofer.pdf
PDF / 5.53 MB / 2024-01-08-Slides-Dieter-Kranzlmueller.pdf
PDF / 5.36 MB / 2024-01-15-Slides-Kaspar-Lebloch.pdf
PDF / 3.06 MB / 2024-01-22-Slides-Franz-Wotawa.pdf