Organizations are making massive investments in artificial intelligence technologies such as Machine Learning (ML). Yet, ML project fail often due to high uncertainty. In this research, we conduct a survey to examine how organizations can cope with the high uncertainty of ML projects. The project is a collaboration between the IT University of Copenhagen and the University of Auckland.What? Why? How?

What? Query optimization in data systems is crucial for efficient query execution. AI/ML have been recently proven very effective tools in enhancing this process, reducing efficiency and manual effort. This project aims to explore the potential of ranking and develop specialized ranking methods to improve query performance. Our vision is for the project to pioneer a new research paradigm. Why? Query optimization emerges as a critical source of optimal data-to-insights time, cost savings and energy efficiency. Despite the potential of ML-based approaches based on regression models, they have yet to achieve optimal performance. In this project, we will investigate what really matters for efficient query optimization, i.e., the relative order of query plans and, thus, ranking algorithms. How? The aim of Rank4QO is to explore ranking methods and devise specialized algorithms to enhance query performance. To achieve this, we will conduct a thorough analysis of the connections between queries and query plans, develop learning-to-rank models, and new optimization algorithms. Ultimately, Rank4QO will reshape the entire query optimization process by incorporating the ranking concept.

What? A lot of research has been devoted to finding quantum algorithms to be run on quantum computers in the future. What has only be realised very recently is that some of the building blocks of these algorithms can also lead to faster classical algorithms to be run on available classical hardware. One area where these quantum-inspired have very recently provided dramatic speedup is in solving partial differential equations. This project aims to develop the full potential of quantum inspired algorithms, and ambitions to dramatically speedup the solution and simulation of partial differential equations in a wide variety of settings. Why? Solving partial differential equations to high precision is essential for many applications in science, engineering and finance, including: chip design, radar cross section, computational fluid dynamics, seismic imaging, quantum chemistry, derivatives pricing in finance, and many more. Globally, a large fraction of high performance computing resources are allocated to solving PDEs. Hence, algorithmic progress will have substantial industrial and scientific impact, while mitigating the CO2 footprint of these massive simulations. How? The essential toolset, upon which the project is built, is tensor networks. Tensor networks were originally developed for simulating strongly correlated quantum materials. They are especially well suited for describing high dimensional objects that contain strong, but ordered, correlations across the dimensions. Solutions to partial differential equations exhibit exactly these sorts of correlations due to their differential origin. Hence, we will leverage a large body of work into a new field, and expect constructive cross-fertilisation.

The VeriFunAI project aims to make artificial intelligence (AI) safer and more reliable by developing new mathematical tools. These tools will help ensure that AI systems work correctly and do not cause harm, contributing to a future where AI technologies are trustworthy and beneficial for society.What? Why? How?

The goal of research in artificial life is to understand and simulate life-like processes using artificial systems. The goal of the International Conference on Artificial Life 2024 in Copenhagen is to share new developments in the field with the aim to make radical new technology that exhibit life-like properties and deepen our understanding of life itself and its origin.What? Why? How?