Tufa Talks episode 1 - Johannes von Oswald (Google)

​We are a small, independent research group working on fundamental AI research.

​Join us for the first in a series of talks hosted at our office. Our goal is to share the latest ML research and foster open discussion and collaboration among ML researchers in Zürich.

​Speaker: Johannes von Oswald (Paradigms of Intelligence Team, Google)
https://jovoswald.github.io/

​Johannes von Oswald is a research scientist at Google Zürich and ETH Zürich PhD alumnus. His work focuses on how neural networks learn from data — particularly on getting transformers to implement algorithmic solutions rather than just pattern match. He is best known for his research on mesa-optimization, showing that gradient descent can emerge within transformer activations.

​Agenda:

​📅 Date: April 28
🕕 Time: 6:00–8:00 PM
📍 Place: Tufa Labs, Nüschelerstrasse 30, 8001 Zürich

​17:45 – Doors open
18:00 – Talk
18:30 – Open Q&A
19:00 – Pizza, drinks, networking

​Please note: doors open at 17:45 and close promptly at 18:00. Make sure to arrive within that window.

​We'll have pizzas and drinks after the talk, hope you can stay!

​Spots are limited, registration is subject to approval.