Understanding the Inner Thoughts of AI — DeepMind on Interpretability 11 July 2026·5 mins Machine Learning Talk Notes Llm Interpretability Alignment Deepmind Mechanistic-Interpretability Sparse-Autoencoders