Anthropic has a strong safety mission, which includes responsible scaling policies to identify and mitigate catastrophic risks associated with AI models.

OpenAI was initially focused on developing superintelligence safely, but over time, safety became less of a priority as they shifted towards becoming a for-profit company.

The founding spirit of OpenAI is focused on solving the safety problem associated with AGI, emphasizing that the greatest minds cannot solve it in isolation; they need to engage with reality and see how technology plays out in practice.

Anthropic is focused on user safety by containing actions within a secure virtual machine and limiting access to sensitive data, although some restrictions may be lifted in the future.

There are opposing views on AI safety: one advocating for open-source transparency to empower users, while the other emphasizes the risks of powerful AI in the wrong hands.

Ilia, one of my students, left OpenAI due to safety concerns, which highlights the seriousness of the risks we face with AI.

There is a delicate balance between being excited about AGI and the apprehension that comes with OpenAI's role in its development, which raises concerns about unilateral decision-making in AI governance.

There's a lot of work to do to ensure AI safety, which encompasses not just avoiding offensive statements but also preventing catastrophic outcomes for humanity.

The original concerns about AI were about existential risks, but the conversation has shifted to societal risks, which often intertwine with political issues.