WP1 – Research on generative AI

Research focus: AI for prevention of white collar crime

Tasks

• Develop a data platform for data generation using generative AI techniques to produce high quality synthetic data.
• Use real-time anomaly detection to highlight activities indicative of money laundering.
• Research detection of generative AI. We will mostly focus on LLMs, but also consider multimodal systems. Further, both white-box and black-box will have a role.
• Dissemination, targeting both the general public as well as NTNU master students.

Potential societal impact

By developing a data platform that generates synthetic data, researchers can share information without exposing sensitive details, promoting faster iterations and innovative research. Existing systems often lack real-time capabilities and suffer from data silos and limited collaboration. New methods using synthetic data can improve the detection of suspicious transactions, preventing financial crimes and reducing risks for financial institutions. Advances in AI, particularly generative AI, can be exploited by fraudsters, making it crucial to stay informed about new threats and educate the public on recognizing manipulation attempts.

WP2 – Responsible AI

Aim: Increasing knowledge on XAI and AI safety, through research activities and dissemination. Create course material and a theory moduleon these findings.

Tasks

• Research the detection of misaligned goals in RL agents
• Developing new and improving existing explainable AI (XAI) methods for misaligment detection
• Research and develop techniques for rectifying inner misalignment in reinforcement learning agents
• Develop and offer a theory module on XAI and inner misalignment, and adapt this to a wider audience

Potential societal impact

This research ensures AI systems operate as intended, making it easier to understand and explain their decisions, which helps prevent biases and protect users. XAI research often focuses on making AI understandable for experts rather than laypeople. As AI becomes more integrated into everyday functions, it's crucial for the public to have a basic understanding to make informed decisions about its use. Developing advanced academic subjects is necessary to drive innovation, support research-based teaching, and produce industry-ready graduates.

WP3 – Innovation Hub

Aim: Develop a more in-depth understanding of how AI-based innovations in industry can be accelerated and supported, with an emphasis on the Midt-Norge region.
 

Tasks

• A yearly AI adoption barometer survey will be run to identify the adoption status and future directions for companies.
• Yearly Innovation Hub workshops with industry participants to present findings and discuss best practices.
• An AI adoption toolbox will be provided to involved companies to aid them in their AI uptake process and present ways to measure and capture business value from their AI investments. 

Potential societal impact

Annual reports summarizing key trends and developments are crucial for guiding collaboration and providing regional ownership of issues. Through iterative processes like evaluations and co-creation workshops, the goal is to generate research-based knowledge to facilitate AI innovation for local businesses. Involving students in this process will enhance their education and improve recruitment opportunities for local businesses. Developing a roadmap for AI services and understanding AI-driven transformation will help inform organizations and decision-makers.