ITEA Topical roadshow 2024

Large Language Models - 26 March 2024

The use of Large Language Models has increased significantly since 2023 as a very interesting technology. As an illustration, the ITEA Call 2023 showed strong interest, with 10 of the 24 Full Project Proposals intending to use LLMs. In today's fast evolving landscape, LLMs offer many opportunities for innovation and business growth. The main challenge is how to use these models effectively to create value.

The 140 registrations for this webinar confirmed the clear interest from the ITEA Community in LLMs. This webinar provided an excellent opportunity for participants to learn more about LLMs, exchange ideas, and collaborate to maximise the potential of this exciting technology.

Topics addressed

• Generative AI for the software development life cycle by Robin Gröpler - Institut für Automation und Kommunikation (IFAK)
• LLMs for new services by Abdelkrim Boujraf - ALT-F1
• Trustworthiness and ethical topics of LLM by Mikko Raatikainen - University of Helsinki
• Generative AI in media: challenges and learnings by Karim Dahdah – VRT

Outcomes

From the different presentations, it is obvious that even if LLMs now represent a mature technology, their integration in business applications needs to be carefully planned. The general-purpose offering is not always able to create business value. The tools need to be tailored to the objectives of your application and some human checks are important. The session also showed that many experiments are ongoing and that the field will continue to progress rapidly. The ITEA research Community is very active in this domain and continued exchange on the topic will result in faster progress thanks to experience and knowledge sharing.

The slides, the full recording and recording of each presentation separately can be found at the dedicated session's webpage.

The legal consequences of using AI in healthcare - 20 June 2024

AI is invading our daily lives, and more AI applications are used every day. This is also the case in the healthcare sector where a high number of applications and tools are developed to support decision-making by medical staff. This current trend is also seen and growing among ITEA projects focusing on Smart Health. In this ITEA Topical Roadshow co-organised with Jan-Marc Verlinden, founder of Medrecord and HealthTalk from the Netherlands, the legal consequences of using AI in healthcare, also in relation to the AI Act, were discussed.

Topics addressed

• eHealth apps and certification by Jan-Marc Verlinden - Medrecord
• Legal: no showstopper by Jos van der Wijst - BG.Legal
• Value creation in healthcare through AI by Milan Petkovic - Philips

Outcomes

During the session, we delved into the AI Act and the associated risks of using AI in healthcare, which remains a grey area for many. The pyramid of risk calculation was explained in more detail and more insights were given into the (categorisation of and requirements for) high risk cases and into what you can or cannot do.

Through various examples, it became clear that many aspects are still ambiguous, leaving room for legal challenges. Consequently, it is advised to ensure accountability and company-wide awareness of obligations, as even developers will be held responsible. Preparation for the AI Act's impact should involve AI mapping, classification, risk assessments and corresponding actions.

In the (near) future, guidelines will come to further clarify the AI Act while standards are being extended and created to ensure compliance.

Finally, the opportunities of using AI in healthcare were presented with the proviso of keeping patient safety as the most important aspect of healthcare.

All presentations and recordings can be found at the dedicated session's webpage.