Learning causal mechanisms

Bernhard Schölkopf
/ Max-Planck-Institut für Intelligente Systeme, Tübingen

24. September 2019
14:00 – 15:30

Abstract

In machine learning, we use data to automatically find dependences in the world, with the goal of predicting future observations. Most machine learning methods build on statistics, but one can also try to go beyond this, assaying causal structures underlying statistical dependences. Can such causal knowledge help prediction in machine learning tasks? We argue that this is indeed the case, due to the fact that causal models are more robust to changes that occur in real world datasets. We discuss implications of causal models for machine learning tasks, focusing on an assumption of ‘independent mechanisms’, and discuss an application in the field of exoplanet discovery.

Smart Cities – Making Cities Livable and Sustainable

Siobhán Clarke
/ Trinity College Dublin, Irland

25. September 2019
9:00 – 10:30

Abstract

Given growing urban populations, it is clear we need to change our behaviour to better manage the sharing of increasingly constrained urban resources, such as the road network, energy, water, and so on. This talk explores how automation, using real-time decision-making, can play a part in assisting citizens in making better use of the resources available to them. The goal is not to take over citizens’ lives, but to remove the onus on citizens to be constantly aware of potential opportunities for optimising resource sharing. In particular, the talk uses examples from autonomous vehicles and energy demand-side management.

GI 50 – und wie geht es weiter?

Volker Claus / Universität Stuttgart
Stefan Jähnichen / Technische Universität Berlin
Reinhard Wilhelm / Universität des Saarlandes

25. September 2019
14:00 – 15:30

Abstract

Als vor 50 Jahren unsere Gesellschaft für Informatik gegründet wurde, standen wissenschaftliche Berechnungen und die Automatisierung von Routineprozessen und somit die Unterstützung von „alltäglichen Prozessen“ im Vordergrund. Doch schon bald eröffneten sich ungeahnte Steigerungen der Leistungsfähigkeit verbunden mit zunehmender Komplexität und undurchschaubaren Auswirkungen. Heute scheint alles, was algorithmisch denkbar ist, auch umsetzbar zu sein. Wie explosionsartig sich die Möglichkeiten bereits entwickelt haben, zeigen nachdrücklich die mehr als 240 Newsletter des GI-Radars – doch das ist wohl nur der Anfang des Einstiegs in eine andere kommende Welt.

Wenn wir für diese andere Welt eine neue GI gründen würden, welche Werte und Ziele würden wir in ihre Satzung schreiben? Wie würden wir Mitglieder werben, welche Regeln und Beschränkungen würden wir für den Einsatz von Algorithmen verlangen, welche Kontrollmechanismen würden wir fixieren, wie professionell würden wir uns aufstellen, welche konkreten Produkte würden wir anbieten, welche politische und wirtschaftliche Bedeutung würden wir für unsere GI einfordern; kurz, wie würden wir eine neue GI (also eine „GI 5.0“), die künftige Aufgaben zu erfüllen hätte, in Deutschland positionieren?

Dieser Frage wollen wir, ausgehend von den sich überall abzeichnenden Einsatzmöglichkeiten von Informatiksystemen, nachgehen und zugleich 50 Jahre GI und ihre aktuellen Aktivitäten betrachten. Was bietet die GI in Zukunft? Wem nützen die Informatiksysteme? Was wollen unsere Mitglieder? Wie soll eine Maschinen-Ethik abgesichert werden? In welche Richtung orientiert sich die Jugend? Wie verändert sich Kommunikation, wie verändert sich der Alltag? Was können und müssen wir zu einer stabilen demokratischen Gesellschaft beitragen? Viele Fragen und Strukturen sind zu diskutieren, um weiterhin mitwirken zu können; denn wie fast alle Institutionen muss auch die GI ihren künftigen Fahrplan, ihre Sichtbarkeit und ihre Einflussnahme den allgemeinen Veränderungen, den Informatik-Innovationen und deren Auswirkungen in der Gesellschaft anpassen.

Data Science for Social Good: Opportunities and Challenges

Ciro Catutto
/ ISI Foundation

26. September 2019
14:00 – 15:30

Abstract

The value of big data and advanced analytics lies critically in the opportunity to make better decisions and to design better policies. Identifying needs, targeting interventions, and measuring impact are all challenges that can greatly benefit from more quantitative approaches and data-intensive methods. This opportunity is currently stimulating new research lines in academia, new data sharing initiatives in industry, and new programs in the non-profit sector, while also calling for novel cross-sector collaborations around data. This talk will reflect on the complex interplay of new data sources, data science methods and algorithmic decisions, discussing selected case studies in the domains of health and mobility, and highlighting opportunities as well as challenges for the generation of public value and social impact.

Computer, Art, and the Artificial

Frieder Nake
/University & University of the Arts, Bremen

25. September 2019
9:00 – 10:30

Abstract

How ever far back into the past I try to remember, one bewildering question keeps re-appearing: “Who really is it who makes the art?” Already before the Gesellschaft für Informatik had even been founded, this question was raised. It was raised when computers were claimed to generate art. Whenever later some extraordinary artistic event attracted the audience’s attention, the more or less same question was asked again, perhaps even more urgently. As, e.g., just recently when the spectacularly high price of $432,000 was paid for an image that was generated by a computer (it happened at an auction at Sotheby’s). The question seems to appear whenever drawings are exhibited in galleries whose process of production involved algorithms, programs, or computers in some novel way. This has been happening since more than fifty years now. Currently, it is happening under the flags of big data or neural networks. – To me, the question always requests the same answer, even though its context may change dramatically. All the time, the answer remains one and the same: it is not the computer that makes the art – an answer that can hardly be considered to be a surprise. In the lecture, I want to discuss this situation by putting it into a critical cultural context. What is it that – from the discourses of aesthetics and art history – we can learn about absurdities and speculations surrounding the computer’s operations, and the results of informatics? I will characterize computers as semiotic engines, humans as semiotic animals (following Felix Hausdorff). The terms both refer to semioses (sign processes). And that’s the reason for the repeated pleasure that people draw from picturing computers as becoming humans when the computer is really nothing but a machine, a machine, it is true, of a special kind, but still a machine, we cannot deny this fact. I cannot avoid from slightly bumping into Artificial Intelligence which, I hope, the audience will tolerate.