Statements

• Broadly slowing responsible frontier labs' AI model and application research would be counterproductive.
• Some current AI agents built around large language models might already have rudimentary subjective experience.