Free event, register below.
Solveig Thrun
Nasjonalt forskningssenter innen komplementær og alternativ medisin, Sykehusvegen
21, 9019 Tromsø, Norway
Can AI Predict Breast Cancer Risk? Advancing Early Detection with Deep Learning
Breast cancer screening saves lives, yet current programs assume all women have the
same risk, overlooking individual differences. I will talk about how deep learning can
analyze mammograms to predict a woman’s individual risk of breast cancer up to 5
years in advance by detecting subtle early patterns. By moving beyond a "one-size-fits-
all" approach, I will show how personalized screening strategies can enable earlier
detection and prevention, ultimately improving patient outcomes.
Daniel Kaiser
Hansine Hansens veg 18, 9019 Tromsø, Norway
CogniLoad: When and why does logical reasoning in LLMs fail?
Deductive reasoning is the cornerstone of any reasoning task that you throw at an LLM.
But how far does it go? In our paper we investigate this question to understand whether
task length, difficulty or distracting information are the bigger issue for current chatbots.
Inspired by Cognitive Load theory we generate logical riddles and ask the most popular
LLMs to solve them to find out how good they are in comparison, and what really
causes them them to fail.