AI Seminar Series

Join Associate Professor Ujwal Gadiraju from TU Delft for this talk discussing the intriguing and pertinent role of human input in propelling better AI technology in the quickly evolving age of generative models.

Join us as we discuss the first comprehensive taxonomy of NFQ categories and the expected structure of answers.

This seminar will show that a differential analysis of language model snapshots before and after an update can reveal a surprising amount of detailed information about changes in the training data. To demonstrate the leakage due to updates of natural language models, we develop two metrics — differential score and differential rank.

This talk will discuss approaches to integrating behind-the-meter distributed energy resources (DER) into low-voltage distribution networks. We approach the problem from two angles: (i) uncoordinated energy management of individual prosumers and (ii) coordinated approaches that orchestrate the response of several prosumers in virtual power plants.

Information warfare instruments have recently been used to weaponise misinformation to foster propaganda and to reach political goals by influencing populations at scale. In this seminar we will discuss how human-in-the-loop AI technology can support expert fact-checking efforts that have been increasing substantially due to the rise in the spread of misinformation.

Artificial intelligence (AI), and machine learning (ML) in particular, are set to revolutionise everyday healthcare and improve patient outcomes. This presentation will illustrate the steps required to take AI/ML applications from in-silico prototypes to effective point-of-care instruments embedded within electronic medical records and imaging software.

Precision prevention of advanced melanoma is fast becoming a realistic prospect, with personalized, holistic risk stratification allowing patients to be directed to an appropriate level of surveillance, ranging from skin self-examinations to regular total body photography with sequential digital dermoscopic imaging.

Prof. Miller will present an overview of the intersection of explainable AI and will present some key examples of how to integrate social science knowledge into these methods for explainability in sequential decision making problems.

Prof John Fraser will explore how we can thrive in the midst of COVID 19 and provide an overview of the global collaboration of CCRG and the future of ICU.

Plant breeders are expected to have a working knowledge of the Breeder’s Equation. It is also expected that the Breeder’s Equation will be used as a framework to help optimise the design of their breeding program. This was relatively straightforward when breeding was based predominantly on selection for trait phenotypes. Today, breeders either use, or aspire to use, genomic information in many ways to improve the effectiveness of their breeding programs.

The issues of food security and sustainable agriculture are vital concerns to society and key topics in assessments of climate variability and change on productivity and food supplies. However, accurate and advance knowledge of the associated risk in crop production systems can mitigate some of the impacts of such causative threats. It is anticipated that such approaches will increasingly become more valuable in decision-making and lead to better preparedness in coping with the impact of extreme climate events leading to a reduction in the downside risk and more resilient agricultural food systems.

Artificial intelligence and the data sciences are enabling ‘Agriculture 4.0’. In 2020, agrifood tech startups raised more than $30 billion in investment. Many of these innovations are deployed in agriculture research, supporting new sustainable solutions that keep us fed. Professor Scott Chapman will discuss some of these digital and predictive agriculture technologies.

Musculoskeletal pain conditions, including low back pain, are the leading cause of disability internationally. Recent work has begun to expose features that have the potential to guide personalisation of treatment, but the challenge to disentangle the complexity of the condition is immense. Artificial intelligence will need to play a role.

Sophia Ananiadou will describe NLP methods for the extraction of structured representations from text for downstream applications while addressing challenges of information extraction tasks.

The concept of "biased data" is often too generic to be useful.  Through a series of cases studies, we will explore what algorithmic bias is, different types (with different causes), and debunk some common misconceptions.  We will cover why algorithmic bias is a problem worth addressing and some steps towards solutions.