Recap of NORA Annual Conference 2024

The conference highlighted Artificial Intelligence's (AI) impact on several fields including biomedicine, misinformation, and ethics. The event featured keynote presentations, thematic tracks, and an AI-centre pitching session with The Research Council of Norway.

Awards were given for various contributions to AI, including Lifetime Achievement, Early Career Investigator, Diversity in AI, Publication of the decade, Startup of the year, and Community Building. The conference included a dinner and award announcements at Kristiansand Zoo and Amusement Park.

 

Gathering the AI-Community in Norway

Professor Morten Goodwin from University of Agder and member of the NORA Board welcomed the 120 participants attending the conference at Clarion Hotel Ernst, in Kristiansand.  

"We are gathered here today as industry professionals and researchers in a time where AI is omnipresent in society" – Morten Goodwin.

The NORA conference is an annual event that aims to gather the Norwegian research community within the field of Artificial Intelligence and create a platform where invited speakers and participants can share research, ideas, theories, models and new perspectives, and interact with peers from the field.

All researchers from NORA Partners and Strategic Partners are invited to participate at the conference. Representatives from industry and the public sector are also welcome.

Knowledge sharing and interaction are at the center of the conference, and at this years’ conference it was evident that we have a strong community of researchers and practitioner in Norway.

Enabling High Performance Medicine using using advances in HPC and AI

Ravi Madduri, Computer Scientist at the Argonne National Laboratory, was the first keynote speaker at the conference. Madduri delivered a talk on the intersection of high-performance computing (HPC) and AI, particularly in biomedicine.

He highlighted the critical role HPC plays in developing large AI models, citing examples such as the use of supercomputers like Summit to analyse extensive genomic data from the Million Veteran Program (MVP).

Madduri discussed the need for collaboration and diversity in data to build more trustworthy and effective AI models. He emphasised the importance of privacy-preserving federated learning, which allows models to be trained on data distributed across different locations without centralising sensitive information.

As researchers and scientists, it is our duty to make sure that the research and AI, especially the models that are being created, are created responsibly and we know what we’re doing and we have some safeguards, some things in place so we can reason why the models are working the way they’re working. - Ravi Madduri

Madduri also touched upon the challenges in AI for biomedicine, including issues of bias, privacy, security, and the need for transparency in model governance. He stressed the necessity of continuous learning and validation of AI models, advocating for randomised clinical trials to ensure reliability. By leveraging advances in applied mathematics and high-performance computing, the goal is to enhance AI’s capability to improve healthcare outcomes while maintaining rigorous standards for privacy and data protection.

Madduri encouraged the audience to collaborate on international federated learning projects to build generalizable AI models that can address diverse medical needs.

Thematic tracks on NeuroAI and AI in Commerce

The conference program was organised into several thematic tracks, each focusing on a critical area of AI research and application. After the opening keynote talk by Madduri, the conference program moved over to the first two thematic tracks:

Track 1 - NeuroAI:

This track delved into the intersection of neuroscience and artificial intelligence, exploring how AI can be used to model brain functions and develop new neurotechnologies. Sessions covered topics such as brain-computer interfaces, cognitive computing, and neuroinformatics.

In this track, the talks included the following presentations:

• Examination of Computational Performance of Biological Neural Networks with Reservoir Computing | Felix Simon Reimers, Ola Huse Ramstad, Axel Sandvig, Ioanna Sandvig and Stefano Nichele

• What’s a Machine? From In-Silico to In-Vitro AI  | Stefano Nichele
• Learning Consistent Causal Abstractions with Genetic Algorithms |  Fabio Massimo Zennaro and Kai Olav Ellefsen
• A scalable, hybrid genetic algorithm for continuous multimodal optimization in moderate dimensions  |  Klaus Johannsen, Nadine Goris, Bjørnar Jensen and Jerry Tjiputra

Track 2 - AI in Commerce & Industry:

Presentations in this track highlighted the role of AI in revolutionising commercial industries. Discussions ranged from the use of AI in customer service and supply chain optimisation to advanced analytics and AI-driven marketing strategies.

In this track, the talks included the following presentations:

• Vibrent: Adapting Recommender Systems to the Fashion Rental Economy  | Karl Audun Borgersen, Morten Goodwin, Morten Grundetjern and Jivitesh Sharma

• Using AI to become the customer’s first choice in an e-commerce checkout  | Ingrid Nyquist og Kathrin Flisnes Bergh
• Multi-step Forecasting in Process Industries  | Pål V. Johnsen, Eivind Bøhn, Sølve Eidnes, Filippo Remonato and Signe Riemer-Sørensen
• Providing Reliable Volume Forecasts using ML and Enabling Optimization in Tactical Planning  |  Sunniva Reiten Bovim and Pål Gudmund Berven.

After the two first tracks were concluded and all the participants had enjoyed the splendid complimentary conference lunch, it was time for the second keynote speaker.

Searching for Structure in Unfalsifiable Claims

Keynote speaker Serge Belongie, the Director of the Danish Pioneer Centre for AI at the University of Copenhagen, presented a project aimed at addressing the spread of misinformation, particularly through unfalsifiable claims on social media.

This project, part of a larger effort termed “moonshots” within the center, seeks to develop hybrid AI-human systems to understand and counteract narratives that go beyond simple fact-checking. Belongie emphasised the complexity and urgency of the issue, especially with the rise of generative AI technologies like ChatGPT and DALL-E, which have intensified the dissemination of misinformation.

“In the big cat and mouse game that we’re talking about, I’m trying to make a bold move so that consumers, regular people, can get more insight into what’s going on with narratives.” – said Serge Belongie.

The project’s approach involves analyzing the underlying narratives in social media posts, memes, and other multimedia content. By focusing on the “vulgar” sense of narratives—those that are implicit and often go unrecognised by the public—the team aims to create a comprehensive system to identify and understand these narratives. This involves a multidisciplinary effort, integrating machine learning techniques like deep metric learning with human annotation, to build a robust and nuanced understanding of how narratives propagate and influence public opinion.

Ultimately, Serge Belongie said, the goal is to create tools, potentially in collaboration with social media companies, that can make these processes transparent and auditable, thereby contributing to a more informed and resilient digital society.

Belongie’s talk was followed up with a lively discussion before the conference program continued with two new thematic tracks.

Thematic tracks on Ethics & Data Privacy and Earth & Geospatial AI

Track 3 - Ethics and Data Privacy

With the growing influence of AI, ethical considerations and data privacy have become paramount. This track featured debates and talks on responsible AI development, ethical AI frameworks, and the challenges of ensuring data privacy in an increasingly connected world.

In this track, the talks included:

• Analysing the Efficacy of Evaluation Metrics for Data Privacy Preservation with Textual Data  | Dan Zhang and Mark Anderson.

• An Explainable AI tool for trustworthy prognosis in acute stroke  |  Luis M. Lopez Ramos, Adam Hilbert, Ayan Chatterjee, Steven Hicks, Michael Riegler, Pål Halvorsen and Dietmar Frey.
• Machine Performance Tradeoffs to Ethical and Legal Perspectives  |  Jim Tørresen, Adel Baselizadeh, Tobias Mahler and Lee A. Bygrave.

Track 4 - Earth and Geospatial AI

This track focused on the application of AI in environmental science and geospatial analysis. Topics included climate modeling, satellite image analysis, and AI-driven approaches to sustainable development.

In this track, the talks included:

• The need for machine learning to break size barriers in studying the behavior of natural lavas and magmas |  Razvan Caracas, Marthe Guren, Tim Bögels, Ana Anzulovic, Anne Davis and Sarah Figowy
• Foundation Models for Earth Observation  | Theodor J. L. Forgaard, Alba Ordoñez, Srishti Gautam, Anders U. Waldeland, Jarle H. Reksten, Michael Kampffmeyer and Arnt B. Salberg
• Geospatial AI in the building permit process | Norkart

Norwegian AI Cloud User Survey

Researcher Martin Holen from University of Agder presented the Norwegian AI Cloud User Survey.  

The Norwegian AI Cloud (NAIC), is a project funded by the Research Council of Norway that aims at providing powerful computational resources for researchers, PhD students, startups, and small to medium-sized businesses. The project involves various Norwegian institutions, including universities and research centers.

Holen discussed the findings of a survey conducted to gather information on users’ needs and resources. The survey revealed that most respondents were technical end users with a strong understanding of AI, primarily from academic backgrounds. The main bottlenecks identified were a lack of computational resources and information on how to use these resources effectively.

Holen highlighted the disparity in computing power between Norway and leading institutions like Oak Ridge’s Frontier server, which has significantly more resources. The survey showed a high demand for GPUs, predominantly Nvidia, and the use of programming languages such as Python, C++, and R, with libraries like PyTorch and TensorFlow being popular. A significant number of respondents manually scheduled their workloads and did not distribute them across multiple servers, indicating a need for better training and tools.

Holen emphasized the importance of addressing these bottlenecks and enhancing resource sharing and scheduling to improve the efficiency of AI research and applications in Norway.

The Research Council of Norway: Norwegian AI Centres - Pitching and Matchmaking

Olaug Råd from the Research Council of Norway, along with colleague Pål Sigurd Malm, facilitated a matchmaking event focused on AI centers and their development. The event aimed to give participants a brief two-minute opportunity to pitch their initiatives. Råd provided background on the AI centers initiative, explaining that many project ideas are currently under consideration. After the deadline on June 7th, the most promising ideas will be developed into full proposals.

The AI centres are intended to generate new knowledge in AI across three tracks: societal consequences, technology, and innovation. Råd clarified that funding from the AI billion initiative is not for infrastructure or computing resources, but there are ongoing efforts to strengthen Norway’s computing capabilities and industry engagement in AI research and innovation. The goal is to eventually establish four to six AI centers, with organisations like NORA playing a crucial role in the process.

The session continued with two-minute pitches from several of the conference participants, followed up with a matchmaking session.

NORA will help coordinate activities

The CEO of NORA, Klas Pettersen, said that NORA, as a national AI network, will help coordinate activities between various AI research centres, create opportunities for networking, strengthen international collaborations, and increase the centres' international visibility.

NORA can also coordinate with its national Research School and connect the centres to NORA’s innovation ecosystem, NORA.startup.

“Through NORA, the Norwegian AI research community is already collaborating, and we believe that stronger national collaboration is crucial. This collaboration is essential for undertaking larger projects of national importance and attracting talent” – Klas Pettersen said.

---

Conference Dinner at Dyreparken i Kristiansand and announcement of NORA Annual Award winners

After a productive first day of the conference, it was time for dinner at the famous Kristiansand Zoo and Amusement Park (Dyreparken).

Kristiansand Zoo and Amusement Park put on a entertaining show.

The show was amazing, but the highlight of the evening was when Professor Stefano Nichele, Director of Østfold AI Hub, announced the NORA Annual Award winners.

Here is a list of the winners:

Lifetime achievement - Arnoldo Frigessi

Arnoldo Frigessi is full professor at the University of Oslo (UiO). He has an outstanding publication record (h-index 46, # citations > 8000, paper with maximum influence: 2420 citations). He is the director of the Oslo Centre for Biostatistics and Epidemiology. He has been the director of one SFF and 2 SFIs, and has supervised more than 40 Phd students. Arnoldo has contributed to important research in AI and has been influential in Norway and internationally through leading research centres and educating students.

Early career investigator - Michael Kampffmeyer

Michael Kampffmeyer is associate professor at The Arctic University of Norway (UiT). He leads the research project Next Generation Explainable Medical Computer Vision (NFR) and has already published a large number of research articles in reputable publishing outlets, with an impressive 3314 citations according to his Google Scholar profile since 2019. He is early in his career, having completed his PhD in 2018, and focuses primarily on Deep Learning and Computer Vision.

Diversity in AI - Leonora Bergsjø

Leonora Bergsjø is associate professor at Østfold University College. She is a prominent scientist in the field of ethical AI, engaged in public discussions around the themes of AI bias and trustworthiness. Leonora is the Chair of NORDE, the Norwegian Council for Digital Ethics.  She is a member of the project ENACT - Ethical Risk Assessment of Artificial Intelligence in Practice (2023-2026). Leonora is a professional councel to authorities, including the Ministry of Education and Research, the Norwegian Digitalisation Agency, the Norwegian Directorate for Children, Youth and Family Affairs, the Directorate for eHealth, the Norwegian Board of Health Supervision, the Norwegian Institute of Public Health, the Ministry of Foreign Affairs, the Ministry of Justice, the Ministry of Defence and the Office of the Prime Minister

Publication of the decade 2013-2023 - Marija Slavkovik

Marija Slavkovik, "Formal Verification of ethical choices in autonomous systems" – For pioneering the formal verification of ethics in AI, an essential yet underexplored area. This paper is published in 2016 in Robotics and Autonomous Systems. It has recently influenced several research works on machine ethics, as recognized by 238 citations on Google scholar.

Startup - DoMore

DoMore Diagnostics is a startup company aiming to transform cancer diagnostics and treatment using artificial intelligence. The company originated from a research project named DoMore, from the Institute for Cancer Genetics and Informatics at Oslo University Hospital. Their technology and scientific results have been published in prestigious medical journals including The Lancet, Nature Reviews Cancer, and Lancet Oncology. Recently, they have been awarded the EIC Accelerator grant, which consists of €2.5 million non-dilutive grant and up to €7.5 million in equity matching from the EIC Fund.

Congratulations to all the winners!

---

Day two of the conference

---

The Wizard of AI

Day two of the conference started with Morten Goodwin introducing keynote speaker Marija Slavkovik, who is the Head of the Department of Information Science and Media Studies at the University of Bergen, and member of the NORA Board.

Marija Slavkovik’s keynote emphasised the importance of machine ethics, especially with the rise of large language models (LLMs). She highlighted AI’s role in serving all people and stressed the need for ethical behavior in AI systems, covering key concerns like de-biasing, explainability, and trustworthiness.

Slavkovik discussed the challenges of automating moral reasoning and the need for democratic input in machine ethics, advocating for real, measurable input from individuals into the morals of machines. She also highlighted the necessity for benchmarks in ethical AI behavior and reflected on previous efforts and shortcomings in developing these benchmarks.

The talk underscored that AI’s potential benefits must be guided by responsible development and ethical considerations to truly serve humanity.

After Slavkovik’s presentation and a QA session, the program continued with three thematic tracks.

Thematic tracks on Validation and Power Law Representations, Transportation and Quantum Computing, and Language Models

Track 5 - Validation and Power Law Representations

The track delved into the critical aspects of AI validation and the intriguing structure of power law representations in neural networks. This session featured four presentations:

• Validating AI-based Scoring Engines: What does the Machine Measure?  | Sima Caspari-Sadeghi
• Examining properness in the external validation of survival models with squared and logarithmic losses  |  John Zobolas, Raphael Sonabend, Lukas Burk, Andreas Bender and Philipp Kopper
• Why Generative AI for Time Series?  |  Erlend Aune, Daesoo Lee and Sara Malacarne
• An analytic Approach towards Understanding the Structure of Power Law Representations in Neural Networks  |  Konstantin Holzhausen, Mikkel Elle Lepperød and Anders Malthe-Sørenssen

Track 6 - Transportation and Quantum Computing

The track explored the intersection of advanced AI techniques in transportation and the burgeoning field of quantum computing.

This session featured four presentations:

• 

  Relevant Object Identification from Qualitative Explainable Graphs in Automated Driving  |  Nassim Belmecheri, Arnaud Gotlieb, Nadjib Lazaar and Helge Spieker
• On Road Traffic Prediction with Graph Neural Networks |  Grunde Wesenberg and Ana Ozaki
• NeRFing the Road: Synthesizing Simulated and Real-world Data for Photorealistic 3D Reconstruction of Large-Scale Autonomous Driving  |  Shaira Tabassum, Gabriel Kiss, Sachin Verma, Kiran Raja and Frank Lindseth
• Quantum Computing and Artificial Intelligence |  Shaukat Ali

Track 7 - Language Models

The track delved into the fascinating world of language models, showcasing advancements and applications that push the boundaries of natural language processing. This session included four presentations:

• PlanBERT: From messy zonal plans to informative vector embeddings |  Henrik Brådland, Morten Goodwin, Per-Arne Andersen and Alexander S. Nossum
• Diagnosis of Alzheimer's Disease through Speech: A Comparative Study of GPT Embedding Models and a Traditional NLP Feature Extraction Method | Arezo Shakeri, Mina Farmanbar, Shaima Ahmad Freja and Yeganeh Hallaj
• Enhancing Transparency in NER Models: A Classification-Based Approach with LIME  |  Morten Grundetjern, Per-Arne Andersen and Morten Goodwin
• Probing Machine Learning Models in Angluin’s Style  |  Ana Ozaki

In addition to these tracks, the conference offered a poster session where researchers presented their latest findings and innovations.

Poster Session

It was a great atmosphere at the poster session with lively discussions and networking.

Explainable AI - focusing on Shapley values and counterfactual explanations

The final keynote presentation was held by Kjersti Aas, the Research Director of the Department of Statistical Analysis and Machine Learning at the Norwegian Computing Center.

Aas presented on the topic of explainable AI, focusing on Shapley values and counterfactual explanations. She highlighted the challenges of understanding complex machine learning models, such as XGBoost, which are often used in applications like credit scoring but are difficult to interpret.

Aas discussed Shapley values, which provide a method for attributing the output of a model to its input features based on cooperative game theory. She explained the computational complexity involved and introduced Kernel SHAP, a method to approximate Shapley values more efficiently.

Additionally, Aas introduced MCCE (Monte Carlo sampling of valid and realistic Counterfactual Explanations), a method developed by her team to generate realistic counterfactuals that handle dependent features better than existing methods. This approach ensures that generated explanations are actionable and valid, addressing the limitations of previous techniques.

Aas highlighted the importance of developing explainable AI methods to provide meaningful insights into automated decision-making processes, ensuring transparency and trust in AI systems.

The conference concluded with an awards ceremony, recognising outstanding contributions to the field of AI, and the announcement of the location for the 2025 conference.

Award for Community building and education

Morten Irgens was honoured with The NORA Award for Community building and education.

The award recognises an individual affiliated with a Norwegian institution who has provided truly exceptional service to the field of Artificial Intelligence by, for example, helping to organise the Norwegian AI community, developing valuable resources for the AI community, initiating and developing initiatives for teaching and educating about Artificial Intelligence, such as courses, research schools, graduate schools, workshops, conferences or seminars (new or existing).

Morten Irgens is the director of innovation at CLAIRE. He has a lot of merits in community building, being one of the driving forces in the establishment of NORA, CLAIRE, Adra, and Norsis. Morten has been on the managerial level in several academic institutions, including OsloMet – Oslo Metropolitan University (previously HiOA), Høyskolen i Gjørvik (now Norwegian University of Science and Technology (NTNU)), Kristiania University College, and Copenhagen Business School.

 

Best Oral Presentation and best poster award

The conference participants voted Stefano Nichele’s presentation, "What's a Machine? From In-Silico to In-Vitro AI" to be the best oral presentation at the conference.

Nourin Mohammad from University of Stavanger was voted as the winner of best poster.

Congratulations!

NORA Annual Conference 2025

Next year we hope to see you all in Halden June 2-3, 2025 at Østfold University College. Bring your friends and colleagues! 

 

THANK YOU!

On behalf of the NORA Secretariat, we would like to thank all the participants and all of you that contributed to make the NORA Annual Conference 2024 a huge success! Thank you!