On June 9–10, we had the pleasure of hosting the annual meeting of the Elisabeth Schiemann Kolleg of the Max Planck Society on behalf of our PI and Schiemann Fellow, Celestine Mendler-Dünner.
Thank you to everyone who joined us for two days of inspiring discussions and exchange!
What if AI predictions and human opinions continuously influence one another?
New work by Wu, Abebe, and Mendler-Dünner shows how AI platforms homogenize opinions where disagreement would otherwise persist.
Talk today 11:30am ET @ #NetSci2026 (PS 2.2).
👉 institute-tue.ellis.eu/en/news/how-...
AI agents are starting to train other AI systems and some are already outperforming human-designed pipelines. 🤖
Maksym Andriushchenko and Ben Rank et al. explore how autonomous agents can optimize LLMs through post-training.
Read the article here: institute-tue.ellis.eu/en/news/teac...
find out more here: institute-tue.ellis.eu/en/news/kyut...
The next step of AI will be physical
This is the focus of @kesai.eu, an open science lab dedicated to the next frontier of physic @kyutai-labs.bsky.social.
KE:SAI brings together leading research in Tübingen and Paris to advance world models and robotics through a fully open technology stack.
At the end of March, researchers from around the world came together in Tübingen for the Conference on Parsimony and Learning (CPAL) 2026. Now, we’re excited to share the official event video 🎥
youtu.be/0ojaAmj-bRg?...
Rio de Janeiro is calling! 🇧🇷
The ELLIS Institute Tübingen is joining ICLR 2026 from April 23–27! 🌎
Will you be there?
👉 Check out all of the Institute's publications for ICLR: institute-tue.ellis.eu/en/news/iclr...
📢 Two Calls from ELIAS - European Lighthouse of AI for Sustainability
Proposals Call: Want to join the ELIAS Node network? Apply by 5 July
🔗 https://bit.ly/4fsvFHG
***
Nominations Call: Know a frontier AI leader? Nominate them for the ELIAS Frontier AI Award by 31 August
🔗 https://bit.ly/4nGkKfQ
The Breach: When AI Starts to Cheat 🚨
New research by PIs Jonas Geiping and Maksym Andriushchenko says we have a major blind spot. Most safety systems plan to rely on a smaller monitor model to watch over an untrusted AI agent.
👉 institute-tue.ellis.eu/en/news/the-...
From Lab to Field: Learn like a human, work like a robot.
That’s the approach behind Tübingen-based startup Polybot for building intelligent robots that operate in fields and greenhouses.
Find the full German text here: institute-tue.ellis.eu/en/news/poly...
This major global event brings together top experts to improve Learning Representations." This is a key part of AI that focuses on how computers organize and make sense of complicated information.
