Walter Isaacson interviews Anthropic co-founders Dario and Daniela Amodei about their complementary roles and the values they say guide the company. They describe their reactions to GPT-2, their departure from OpenAI and their approach to training Claude using broad principles.
The conversation explores tests for deception and sycophancy, interpretability research and Anthropic’s stated concerns about autonomous weapons and accountability. The Amodeis also discuss brain-computer interfaces and how people might find meaning if AI changes the nature of work. Their predictions and claims about company practices are presented as their own views.
Watch on YouTube
