1st Industry Collaboration and Transfer Exchange on AI – “From Research to Market”

1st Industry Collaboration and Transfer Exchange on AI – “From Research to Market”

6-7 May 2024
Venue: Innovation Center Saarland University | Saarbrücken Campus, Germany

Welcome to the inaugural Industry Collaboration and Transfer Exchange on AI – “From Research to Market”! This revolutionary congress is dedicated to catalyzing collaboration and streamlining the transfer of knowledge and expertise from industry to academia in the dynamic field of Trustworthy Artificial Intelligence.

This event – co-organised with the support of VISION – is designed to primarily bridge the gap between industry needs and research capabilities. Immerse yourself in inspiring keynote speeches and engaging workshops that aim to define requirements and cultivate micro-projects in close partnership with research entities.

This congress serves as a pivotal platform for industry professionals to communicate their insights, challenges, and requirements directly with researchers. This direct interaction enables researchers to effectively tailor their efforts to address real-world problems and market demands. Simultaneously, researchers get the opportunity to showcase their latest advancements and collaborate with industry partners to ensure their work translates seamlessly into practical solutions.

Beyond the structured sessions, our congress offers extensive networking opportunities and informal discussions. Participants can forge meaningful connections, exploring potential collaborations extending far beyond the event itself.

Embark on this transformative journey with us as we facilitate the seamless transfer of knowledge and innovation from industry to academia, shaping the future of Artificial Intelligence research and its practical applications in the market. Join us in redefining the landscape of AI collaboration and contributing to the evolution of cutting-edge research into tangible market solutions.

Workshops | Parallel Working Groups:

Workshop 1: Bleeding edge innovations or reliable solutions?
The suitability of research results for industry transfer depends on their Technology Readiness Level (TRL), ranging from cutting-edge (TRL8) to moderately advanced (TRL5). While TRL8 findings are readily adoptable, TRL5 results can be tailored to align closely with industry requirements. Hence, the key considerations are identifying the TRL of research outcomes and understanding the industry partners’ specific needs. Let’s discuss these aspects and the existing achievements together.
Workshop 2: Reliable and Efficient Generative AI (GenAI)
Join us in discussing the future of GenAI and its on-device & in-cloud applications. Efficient fine-tuning proves its effectiveness against long and extensive training. Consequently, forthcoming hybrid AI models will have the ability to operate both locally on devices and in the cloud. Decisions about which model to employ will be influenced by factors such as 5G/6G network latencies, the availability of computational resources, and legal considerations.
Workshop 3: Robustness of AI/ML Models
Robust AI/ML models are designed to maintain validity, performance, and resilience throughout their lifecycle, even in the face of changes or attacks. This involves implementing strategies to continuously monitor and adapt models to evolving conditions and potential threats. By prioritising robustness, organisations can enhance the reliability and effectiveness of their AI/ML solutions.
Workshop 4: Trustworthiness of AI systems
Join us to explore how user needs influence the effectiveness of explanations versus analysability in promoting trust in AI/ML models, particularly in specialised contexts, such as neural network-based control systems. In some cases, such as those involving process engineers, analysability (e.g., sensitivity analysis), may be more beneficial than explanations. Tailoring approaches to provide the most relevant insights for users’ specific needs can ultimately increase confidence in AI/ML models.
Workshop 5: Bridging Simulation and Real Measurements
When developing AI/ML/statistical models, the gap between simulation (surrogate models) and real measurements must be closed – both sources of data are necessary, as typically either one alone lacks sufficient quantity or accuracy. How can these be closed?
Workshop 6: Data Quality and Accessibility
Discover with us the growing acknowledgement of the importance of prioritising data quality, accessibility, and regulatory measures alongside algorithms to foster trustworthy AI. Initiatives like the European Commission’s Data Governance Act and Common European Data Spaces aim to enhance data access. Furthermore, explore the potential of synthetic data as a promising alternative for AI applications, particularly when real data access is limited, aligning with efforts to improve trustworthiness and facilitate research and industry transfer in AI/ML.
Workshop 7: AI in business processes and hidden risks
AI can streamline existing business processes, delivering immediate efficiency gains and simple return on investment calculations. Conversely, AI can also revolutionise non-existent or fundamentally changing processes, which is a riskier, but potentially more impactful route. Ultimately, the industry’s appetite for risk will determine its approach to adopting AI solutions. But how high is it?
Workshop 8: AI models and privacy
AI models applied in upstream activities rely on sensitive data, such as individual load curves and contract details, necessitating privacy safeguards to align with GDPR and national regulations. This involves implementing privacy-preserving practices like Differential Privacy in Stochastic Gradient Descent and defending against privacy attacks, enabling scalability across diverse applications and data types, including consumer or IoT data and structured or unstructured data. Discuss with us the necessary measures and their effective implementation to enable scalability across different data types and applications.
Workshop 9: Neuro-symbolic AI: expert augmented AI
Multiple industry organisations use expert systems that include rules, ontologies and knowledge graphs for downstream activities, alongside neural models for learning from data. Combining symbolic reasoning with neural approaches improves performance, provides explainability and increases confidence in the model results, especially when symbolic knowledge cannot be directly converted into physical equations for neural networks. The aim of the workshop is to share methods and experiences in integrating symbolic and neural thinking, especially in scenarios where symbolic knowledge cannot be easily translated into neural networks.
Workshop 10: TrustedAI: Generative AI to empower Traffic Event Scenario Generator
Join us in discussions while tackling the issue of out-of-distribution (OOD) data by leveraging generative AI and constraint satisfaction problem (ML) techniques, enhanced by a collection of explicit (symbolic) attributes. Our aim is to generate in-distribution traffic scenarios, including corner/critical traffic events, to achieve automation levels L3-L5 (SAE). Through a transfer of knowledge and know-how, delve into the significance of these challenges and the proposed solutions for the industry.

Program:

6 May 2024

12.30 – 13.30
Welcome & Reception

13.30 – 13.45
Synergy for Innovation: Enhancing Research Commercialisation Through Corporate and Startup Collaboration
Thomas Neubert – Founder/MD Transatlantic AI eXchange (TAIX)

13.45 – 14.05
European AI Talent Networks bridging Academia and Industry

Laure Poirson – EIT ICT Labs Germany – AI Grid Project, Mohamed Behery – CLAIRE | Rising Researchers Network, RWTH Aachen University

14.05 – 16.00
Parallel Working Groups

16.00 – 16.30
Coffee Break

16.30 – 18.00
Parallel Working Groups

18.30 – 21.00
Dinner & Socialising

7 May 2024

08.30 – 09.00
Welcome & Reception

09.00 – 09.15
Trustworthy AI as a Driving Force for Societally Beneficial Innovations
Kevin Baum, German-Research Center for Artificial Intelligence (DFKI)

09.15 – 09.35
Best practices of Academic-Industry Collaboration from the VISION Project

Elizabeth EL HADDAD, Inria, Beatrice Bozzao, Intellera Consulting S.p.A.

09.35 – 12.30
Parallel Working Groups

12.30 – 13.30
Lunch

13.30 – 15.30
Parallel Working Groups

15.30 – 16.30
Plenary presentation of key findings from the Parallel Working Groups

16.30
Closing

Published
Categorized as Event