MCML Munich AI Day 2025

1:00 pm – 2:00 pm

Registration opens with cold drinks and poster session.

This poster session is part of the ELLIS Pre-ICML 2025 Fest.

2:00 pm – 2:15 pm

Opening of the MCML Munich AI Day: Short greeting by Eva Schulz

2:15 pm – 2:45 pm

Keynote - Julia Schnabel (MCML / TUM)

Seeing the unseen: the role of AI in precision radiology

Abstract:
Typically, radiologists are trained on a limited number of cases in a single hospital, and may not be sufficiently exposed to imaging of rare diseases, early signs of disease, nor incidental unexpected findings.
Variability in patient anatomy and physiology, image quality and disease presentation, and the increasing radiological workload, provide them with yet further challenges.With the shift from diagnostic imaging to screening for early disease detection and patient prognostics, AI is playing an active role in advancing the field of precision radiology. This ranges from image quality control, over screening images for potential anomalies, to generating further imaging on demand, taking AI in a full radiological cycle for both patient and clinician benefit.

2:45 pm – 3:15 pm

Keynote – Tina Eliassi-Rad (Northeastern University)

The Trilemma of Truth in Large Language Models

Abstract:
We often attribute human characteristics to large language models (LLMs), claiming they “know” certain things. However, LLMs have internal probabilistic knowledge. To assess their knowledge, we need to probe them. In this talk, I will present sAwMIL, an approach for separating true, false, and unverifiable statements in LLMs. sAwMIL addresses flawed assumptions in previous work and outperforms other probing approaches on several criteria such as correlation, generalization, selectivity, manipulation, and locality. I will showcase the results of a study involving 16 open-source LLMs and three new datasets. If time permits, I’ll outline my research program on measuring epistemic instability in human-AI society. This is joint work with Germans Savcisens.

3:15 pm – 3:45 pm

Coffee Break & Poster Session

This poster session is part of the ELLIS Pre-ICML 2025 Fest.

3:45 pm – 4:15 pm

Keynote – Neil Lawrence (University of Cambridge)

The Atomic Human: Being Unique in the Age of AI

Abstract:
Our fascination with AI stems from the perceived uniqueness of human intelligence. We believe it’s what differentiates us. Fears of AI not only concern how it invades our digital lives, but also the implied threat of an intelligence that displaces us from our position at the centre of the world.
In this talk Neil will introduce the notion of the Atomic Human, the indivisible component of our humanity that can’t be taken by the machine. He will argue that the algorithmic decision making that emerges from the machine offers a different place to stand and better understand our own intelligence and what is precious about us. Without this perspective we do risk displacing our human parts in favour of the machine, but by seeing our intelligence from the machine’s perspective we can ensure that we integrate these new technologies in ways that enhance who we are rather than replace who we are.

4:15 pm – 4:45 pm

Keynote - Björn Ommer (MCML / LMU)

Generative AI and the Future of Intelligence

Very short abstract:
Generative AI is becoming a critical enabling technology for the development of future technologies. It turns intelligence into a commodity, thereby fundamentally changing information processing and our use of computers. We will then discuss recent developments in AI and their implications on society.

4:45 pm – 5:15 pm

Coffee Break & Poster Session

This poster session is part of the ELLIS Pre-ICML 2025 Fest.

5:15 pm – 6:00 pm

Science Panel

LLMs in Scientific Practice and the Social Consequences

Confirmed panelists

• Mennatallah El-Assady (ETH Zurich),

• Frauke Kreuter (MCML / LMU)

• Neil Lawrence (Cambridge)

• Daniel Cremers (MCML / TUM)

6:00 pm – 8:00 pm