· Event Recap
Speaker Series Summer 2026: seven talks to rewatch
From training LLMs on thousands of GPUs to neural digital twins – seven researchers from Hugging Face, Cohere Labs, Oxford, ETH Zurich and Stanford. All talks are on YouTube.
From April to July, we hosted a talk every two weeks at TU Berlin. The topics went from LLM training at scale and multilingual models to human agency, synthetic biology, vision models and AI safety.
On 14 July, Sophia Sanborn (Stanford University) gave the last talk of the semester. She showed how neural digital twins, deep learning models that predict the responses of real neurons, let researchers run neuroscience experiments in silico.
All seven recordings are on our YouTube channel:
- Nouamane Tazi (Hugging Face): Scaling LLM Training to Thousands of GPUs
- Marzieh Fadaee (Cohere Labs): Scaling LLMs Across Languages
- Christopher Summerfield (University of Oxford): AI and Human Agency
- Georgia Channing (Hugging Face): When Silicon Meets Carbon: Bringing Synthetic Biology to Life
- Thomas Wimmer (ETH Zurich): Beyond Patches: Learning Dense Visual Features
- Fazl Barez (University of Oxford): Towards Automated Interpretability for AI Safety
- Sophia Sanborn (Stanford University): Virtual Neuroscience with Neural Digital Twins
Thank you to all speakers and to everyone who came along this semester! Abstracts and photos of all past talks are on the Speaker Series page.
