An interview Stanford PhD candidate Alex Tamkin, whose research focuses on understanding, building, and controlling pretrained models, especially in domain-general or multimodal settings