NCSR Demokritos Summer School 2026: AI & Business Innovation and Explainable AI

For another consecutive year, NCSR "Demokritos," Greece's largest research centre, held its Summer School, running 6–10 July 2026 in Athens. Now in its 61st edition, the programme gave students passionate about science and research the chance to explore the centre's research potential, with lectures in the main amphitheatre spanning biotechnology, neuroscience, machine learning and quantum computing, alongside dedicated small group talks and a poster session for more specialized interests.

At the Central Amphitheatre, Dr.Denia Kanellopoulou, delivered a talk on AI & Business Innovation, part of the Summer School's DeepTech & Innovation track. The talk drew on interviews with over 200 companies, mainly in Greece, conducted between 2020 and 2022. Using Gartner's AI Maturity Model, the research found that 67% of Greek companies surveyed remained at the "awareness" stage, only 6% had reached "operational" AI use, and none had fully embedded AI into their organisational DNA. She presented APSS (Awareness, Piloting, Scaling, Sustainability), a practical adoption roadmap developed and field-tested through the team's work , with real-world applications drawn from Smart Attica, Pharos and the QUALCO Fellowship. She closed by arguing that in the European context, AI regulation, including the AI Act, tends to act as a catalyst for trust-based adoption rather than a barrier to innovation.

Earlier the same day, Dr. Kanellopoulou led one of four parallel tracks at the 4th Speed Mentoring Session. The format gave students direct, one-on-one access to researchers across disciplines to discuss career paths and professional development.

At the poster exhibition, researchers presented their latest work to Summer School visitors. Thodoris Lymperopoulos showcased his research  in Explainable AI, introducing a novel attribution method for estimating feature importance in a model's decision making process. His approach relies on an evolutionary score that captures the model's attention at each training step, illustrated through a biological analogy: just as evolutionary phenomena are best understood as they occur, rather than reconstructed millions of years later, his method offers real-time insight into model behaviour as it trains. The framing made the concept accessible to a broad audience of biologists, chemists and computer scientists alike.

Read more:  Summer School Demokritos 2026