Presentation from lunch seminar, May 9th:
Modeling and Engineering Beyond Narrow AI:
Link to the presentation (pdf) >>
Dr. Bernhard A. Moser, Research Director, Software Competence Center Hagenberg GmbH, Hagenberg, Austria
Today’s machine intelligence is driven by the Big Data paradigm. Despite its success in recent years, there are severe limitations for a broader scope of applicability. We outline two directions beyond these limitations: First, there is the problem of limited data, which requires transfer learning to compensate for the lack of data by extracting knowledge from trained models in a similar but different setting. Second, there is the problem of limited semantics and contextualization of a pure statistical approach which calls for integrating linked data and relational machine learning. In this context we discuss application scenarios, challenges and progress from ongoing research at SCCH.
References:
- Fischer, L. Ehrlinger, V. Geist, R. Ramler, F. Sobiezky, W. Zellinger, D. Brunner, M. Kumar, and B. A. Moser, “AI System Engineering—Key Challenges and Lessons Learned.” Machine Learning and Knowledge Extraction 3, pp. 56-83, 2021, https://doi.org/10.3390/make3010004
- Zellinger, N. Shepeleva, M.-C. Dinu, H. Eghbal-zadeh, H. D. Nguyen, B. Nessler, S. Pereverzyev, B. A. Moser, “The balancing principle for parameter choice in distance-regularized domain adaptation”, 35th Conf. on Neural Information Processing Systems, NeurIPS 2021, see also www.S3AI.at.
- Haindl, G. Buchgeher, M. Khan, B. A Moser, “Towards a Reference Software Architecture for Human-AI Teaming in Smart Manufacturing”, The 44th Int. Conf. on Software Engineering, ICSE-NIER 2022, https://arxiv.org/abs/2201.04876, see also www.teamingai-project.eu