In the domains of aeronautics, automotive, energy, manufacturing and retail, Munich DIH proposes novel solutions to counter the complexity and dependability challenges resulting from distributed accountability, the need for more efficient and intuitive human-CPS interactions as well as the speed and robustness of constantly evolving AI systems.
In the experiment, a proposal for the deployment of the platform for intuitive, accountable AI (PIANAI) for AI-based systems is developed based on clear AI-based innovation projects in these domains, and implemented as an overlay of the DIHIWARE and AI4EU platforms. Moreover, PIANAI services will be aligned with the AI Manufacturing Testbed in regard to International Data Spaces (IDS), GAIA-X and federated learning. In particular, corresponding interfaces will be added and integrate PIANAI services into the testbed. One application could be the implementation of the “clearing house” concept as defined by IDS.
In summary, the experiment offers a technical service as it provides technical means for the definition of verifiable claims regarding the design, deployment, and consumption of an AI service. The claims are defined and supported through tamper-proof facts along the developing process for the AI service under study. Additionally, the claims can be defined post design and development of an AI service, by handling a number of guarantees regarding the service from a Blackbox perspective
The experiment – as a bundle of “small” experiments” – contributes with PIANAI as a technology service to the DIH4AI platform and benefit from the latter as follows.
First, regarding the DIH4AI platform as an on-demand platforms for innovative AI solutions “made in Europe”, the experiment already integrates #4 regional AI Assets.
Then, regarding the envisioned regional-European platforms interoperability framework, the experiment focuses on analyzing and designing functionalities along IDS and GAIA-X principles towards the integration of the regional platform PIANAI integrated. This effort includes the topic of cloud-edge interoperability in terms of containerized services provided.
Additionally, the experiment contributes to the objective of creating Innovation and Collaboration Platform for DIHs in terms of providing #2 Assets to be accessed by the DIH4AI platform: the PIANAI service as well as a federated learning service.
The experiment also serves as a basis for operationalizing AI Applications Ethical Assessment and Certification by assessing #4 AI applications and providing verifiable claims for their auditing, with the Intra-DIH Regional experiment also involving #2 SMEs and potentially engage up to #4 in follow-up activities.
Finally, the experiment foresees cross-DIH collaborative activities, where a joint provision and development are planned to take place together.