The biggest contradiction about Anthropic CEO Dario Amodei is right there in his job: he believes AI could destroy the world, and he co-founded one of the most aggressive frontier AI labs on the planet.
The extremes were on display long before Anthropic existed. During his OpenAI days, Amodei's caution reportedly bordered on the absurd. Sensitive memos never touched Google Docs. Instead, he'd write them at home on a completely offline computer, then print them out to hand to colleagues. Some people who worked with him say he even refused to travel to China, worried he'd be kidnapped.
Then there's GPT-3. Before the model had even started training, Amodei was apparently concerned it might already be approaching AGI. His safety team reportedly made sure Microsoft was worried about it too.