Results
Content
AI Framework "OPACA"
OPACA Logo
The development of distributed systems is very complex and presents developers with numerous recurring tasks. In our AI framework "OPACA", we combine aspects of multi-agent systems with microservices and container technologies to simplify the development, deployment and linking of new types of applications. In addition to a standardized interface (API), the OPACA platform offers various basic functionalities that are automatically available to all applications developed with it.
- Reference implementation https://github.com/gt-arc/opaca-core
- B. Acar et al., „OPACA: Toward an Open, Language- and Platform-Independent API for Containerized Agents,“ in IEEE Access, vol. 12, pp. 10012-10022, 2024, doi: 10.1109/ACCESS.2024.3353613.
XAI Method Cards
XAI Method Cards (Selection, Front and Back)
Information on methods for explainable artificial intelligence (XAI) is usually only available in complicated and highly technical scientific papers. Reading and understanding these requires a certain level of prior experience and time. For this reason, we have developed profiles of well-known XAI methods, which present information such as possible uses, advantages and disadvantages of the individual methods in simple language. In addition, the XAI methods have labels that make it easier to select them.
XAI Web Demo
Screenshot of "LaFAM" XAI Demo
Artificial intelligence often appears to be an opaque "black box" whose decisions are incomprehensible. Why did the AI make this decision? How "sure" was it? What would have been the next best answer? XAI can help here, for example by highlighting the areas of an image that led - or would have led - to a particular decision.
- Live-Demo in the browser: https://gt-arc.github.io/go-ki-demo/
XAI Tutorial & Method-Finder
Screenshots of the XAI Method-Finder
There are a large number of different XAI methods. This oversupply makes it difficult to find a suitable XAI method that fits a specific use case and is also suitable for the respective target group. We have therefore created two tools to help with this. The XAI Tutorial offers a good introduction for newcomers to XAI to show the possibilities that XAI can offer for an AI application. The XAI Method Finder then helps everyone to find a suitable XAI solution for their specific use case.
- Live-Demo in the browser: https://gt-arc.github.io/go-ki-demo/
Online course "My chatbot"
Screenshot of the Chatbot course
Ever since the emergence of large language models, everyone is talking about Chatbots, but for many they remain a mystery. In our interactive chatbot course, we playfully guide interested users through the theoretical and technical basics of chatbot development, guided by - of course - a chatbot!
- Online Chatbot course http://chatbot-kurs.irml.dai-labor.de/
Go-KI Paper Reading Group
Impressions from the first Meetup in February of 2024
Since the beginning of 2024, we have been organizing regular meetups in the form of a paper reading group at ZEKI (Center for Tangible Artificial Intelligence) in Berlin. We present various papers that are thematically related to the Go-KI project, from the latest findings in the field of (Explainable) Artificial Intelligence to new trends in software development. The meetings will be hybrid (on-site and via Zoom) and are open to all interested participants.
- Registration via Go-KI.org https://go-ki.org/anmeldungen/
- Group on Meetup.com https://www.meetup.com/de-DE/berlin-tangible-xai-group/
- Recordings on Youtube https://www.youtube.com/channel/UCms0AZBeKcwDobZmkTkgaMw
Publications
- Benjamin Acar, Tobias Küster, Oskar F. Kupke, Robert K. Strehlow, Marc G. Augusto, Fikret Sivrikaya and Sahin Albayrak: OPACA: Toward an Open, Language- and Platform-Independent API for Containerized Agents. IEEE Access, vol. 12, pp. 10012-10022, 2024, doi: 10.1109/ACCESS.2024.3353613.
Benjamin Acar, Martin Berger, Marc G. Augusto, Tobias Küster, Fikret Sivrikaya and Sahin Albayrak: An Agent-based Data Acquisition Pipeline for Image Data. IEEE Access, vol. 12, pp 102440 – 102448, 2024, doi: 10.1109/ACCESS.2024.3431429
Aray Karjauv and Sahin Albayrak: LaFAM: Unsupervised Feature Attribution with Label-free Activation Maps. German Conference on Artificial Intelligence (Künstliche Intelligenz), 2024
Tobias Schulz, Pedram Babakhani, Moritz Heine, Finn Harms, Marc G. Augusto and Sahin Albayrak: GGAS: Ground Guiding Assistant System for Interactive Indoor Wayfinding. 14th Int. Conf. on Indoor Positioning and Indoor Navigation (IPIN 2024)
Nils Ole Breuer and Sahin Albayrak: Towards Automated Human-Centered
Recommendation of Explainable AI Solutions. Workshop on Multimodal, Affective, and Interactive Explainable AI @ European Conference on Artificial Intelligence- Abdullah Kiwan, Andreas Lommatzsch and Sahin Albayrak: Enhancing Data Accessibility: Integrating Speech-Based Interfaces with Large Language Models for Intuitive Database Queries. LWDA Workshops 2024
