I had a lot of fun chatting with Shane Legg - Founder and Chief AGI Scientist, Google DeepMind! We discuss: Why he expects AGI around 2028 How to align superhuman models What new architectures needed for AGI Has Deepmind sped up capabilities or safety more?