The authors argue that sophisticated Large Language Models (LLMs) like ChatGPT are full-blown linguistic and cognitive agents. They defend this position by starting with high-level observations of LLM behavior, such as answering questions and making suggestions, rather than focusing on low-level computational details. The authors employ 'Holistic Network Assumptions' to argue that these models possess understanding, beliefs, desires, knowledge, and intentions. They address various objections based on LLM failures, arguing that these do not preclude agency. --- Why it matters: This matters because it challenges the prevailing methodologies in AI philosophy and provides a new perspective on the nature of cognition in artificial intelligence systems. It has implications for how we understand and design intelligent machines. Source: https://arxiv.org/abs/2504.13988