Artificial Intelligence Applications and Innovations. AIAI 2023 IFIP WG 12.5 International Workshops

Ilias Maglogiannis, Lazaros Iliadis, Antoniοs Papaleonidas, Ioannis P. ChochliourosIFIP 2023 · DOI · OpenAlex

The 19th Artificial Intelligence Applications and Innovations (AIAI 2023) conference served as a pivotal forum for bridging the gap between theoretical AI models and robust, real-world implementations. The associated workshops focused on critical sectors including telecommunications, energy, healthcare, and ethics.

The editors of the AIAI 2023 workshop proceedings and their respective affiliations.

Addressing the "Deployment Gap"

A central theme across the proceedings is the "deployment gap"—the challenge of transitioning AI from controlled lab environments to complex, high-stakes real-world systems. The editors identify four primary hurdles:

  • Opacity: The inherent "black box" nature of deep learning limits its adoption in safety-critical fields like aviation and medicine.
  • Efficiency: The requirement for low-latency AI at the network edge to support 5G/6G and robotics.
  • Sustainability: The need for AI-driven management of decentralized green energy grids.
  • Ethics: Navigating the rapid rise of Generative AI and its impact on education and bias.

Methodological Shifts

To overcome these hurdles, the workshops highlighted several novel approaches:

  • Interpretable-By-Design Learning: Shifting from post-hoc explanations to models that are inherently transparent, such as prototype-based systems.
  • Evolutionary Neural Architecture Search (NAS): Automating the design of efficient neural networks for privacy-sensitive federated learning environments.
  • Visual Analytics (VA): Integrating machine learning with interactive visualization to allow human experts to guide AI reasoning.
  • Edge Intelligence: Utilizing 5G network "slicing" to provide dedicated AI processing for autonomous systems.

Diverse Data and Domain Applications

The research leveraged a wide spectrum of datasets across various domains:

  • Biomedical: Ranging from fetal ECG recordings to scRNA-Seq data for Alzheimer’s research and medical imaging for cancer detection.
  • IoT & Engineering: Utilizing sensor data for smart building energy management and water quality monitoring.
  • Social/Humanistic: Analyzing Greek-language social media for geoparsing and Wikipedia for readability.

Key Insights and Results

The workshops emphasized that high-stakes fields are moving away from uninterpretable models toward "human-in-the-loop" systems. In energy, the focus has shifted to proactive management of local micro-grids to meet climate goals.

The rigorous selection process resulted in a 43% acceptance rate (37 of 86 submissions). Notable outcomes include:

  • 5G-PINE: Successful network slicing for robotics.
  • AIBMG: Real-time thermal comfort and self-protecting IoT gateways.
  • MHDW: Advanced ensemble learning for medical diagnostics and sentiment analysis.
  • VAA-CP-EB: Improved energy disaggregation using conditioned autoencoders.

Future Directions

Ongoing challenges include the need for regulatory frameworks for Generative AI, improving the adversarial robustness of automatically designed networks, and balancing the benefits of mobile sensing with individual data privacy.

Summary updated

BibTeX
@inproceedings{maglogiannis2023artificial,
  author = {Ilias Maglogiannis and Lazaros Iliadis and Antoniοs Papaleonidas and Ioannis P. Chochliouros},
  title = {{Artificial Intelligence Applications and Innovations. AIAI 2023 IFIP WG 12.5 International Workshops}},
  booktitle = {IFIP},
  year = {2023},
  doi = {10.1007/978-3-031-34171-7},
}
Figure