Recent & Upcoming Talks

Myth: AI will save us from climate change

AI provides powerful tools to tackle climate change in various applications – but it is not a silver bullet. It can support the mitigation of climate change, for instance, by helping reduce greenhouse gas emissions within various applications. It can support adapting to a changing climate. AI can even support climate science itself. However, it can also be used to harm the climate. To avoid that, AI applications should be developed in collaboration and ongoing exchange with the communities that will use or are otherwise affected by the technology to avoid unforeseen impacts and drawbacks.

Towards a comprehensive assessment framework for the implementation of sustainable AI systems

This expanding use of AI systems raises discussions about their risks, such as non-transparent decision-making processes, discrimination, increasing energy consumption and greenhouse gas emissions or increasing consumption, e.g., through personalised advertising. This talk presents preliminary insights from the project “SustAIn – Sustainability Index for Artificial Intelligence”, which aims at developing a comprehensive framework for assessing AI systems from a sustainability perspective. We outline conceptual considerations for establishing a comprehensive set of criteria that can be deployed in order to minimise negative social, ecological and economic impacts of AI-Systems and their applications. To develop a comprehensive assessment framework, we distinguish between the different scopes of impacts related to AI and we define sustainable AI maturity levels. We outline how we can define measures for each of the above criteria that need to be implemented to enhance the sustainability of AI systems.

Neue Berliner Luft - Laternenladen für Berlin und Chance der Sektorenkopplung in Quartieren

Vorstellung des Projektes neue Berliner Luft.

Using Mobility Data to Simulate the Impact and Opportunities of Electric Vehicles in the Smart Grid

The Open Online Data Meetup (OODM) is an online-based meetup series which provides the space to share interesting insights and entertaining stories from the field of data science with other people from the community. OODM is organized by the CorrelAid Education Team together with CorrelAidX Bremen.

Datenanalyse von Haushalts und Gebäudelastprofilen - Distanzmaße, Prognosefehler und Mittelwerte im Kontext von Smart Meter Daten im Niederspannungsnetz

Wir zeigen, dass klassische Analysemethoden, die auf der Euklidischen Distanz und dem Arithmetischen Mittel beruhen für die Analyse von Smart Meter Daten unter gewissen Umständen zu ungewünschten Ergebnissen führen. Für Smart Meter Daten wurde daher die LPI-Distanz vorgestellt. In WindNODE wurden Algorithmen entwickelt, die die LPI-Distanz für Prognosen oder Cluster-Methoden nutzbar machen. Dabei müssen die unterschiedlichen Ergebnisse in der Datenanalyse im Kontext des eigentlichen Problems evaluiert werden. So zeigen wir, dass die Wahl der „besten“ Prognose vom Optimierungsziel abhängt.