Dario Amoedi worries that AI is developing too fast for us to control risks. He says Anthropic will start hosting independent auditors and that other AI firms should be required to do the same. Seems like a good idea, but I hope they will be more independent than home inspectors or organic-farming certifiers that get paid by those they certify. I also think AI data centers should be required to have physical systems that would allow emergency services (but not random activists) to disconnect from the internet and from power, including any backup power. I use AI for internet searches and for help with Python, but if the experts think AI’s could cause human extinction, who am I to disagree?