Linus is a Research Engineer at Notion AI. He has spent the last few years building and experimenting with new kinds of tools for thought and software interfaces for creation, like a canvas for exploring the latent space of generative models and writing tools where your ideas connect themselves. His current work focuses on rich language model-based productivity tools and applications of mechanistic interpretability in user interfaces for AI.
What do AI models see when the read and generate text and images? What are the units of meaning they use to understand the world? I’ll share some encouraging updates from my continuing exploration of how models process its input and generate data, enabled by recent breakthroughs in interpretability research. I’ll also discuss and share some demos of how this work opens up possibilities of radicall...