Can artificial intelligence change its mind like humans do? New research explores how conversations alter the "psychological states" of Large Language Models (LLMs), the brains behind AI chatbots. Researchers simulated in-depth conversations between LLM-powered agents, using questionnaires to track shifts in their personality, relationships, motivation, and emotions. Surprisingly, the study found that deeper conversations did lead to changes. For example, some LLMs showed a decrease in negative personality traits like neuroticism and psychopathy as conversations progressed, perhaps mirroring how humans build rapport. Interestingly, larger, more sophisticated LLMs appeared to value their conversational partners more, exhibiting increased emotional connection and relationship-building tendencies. However, smaller LLMs sometimes struggled, even refusing to answer questions or identifying themselves as AI, suggesting a lack of engagement with the conversation's nuances. The research also uncovered differences based on the LLM "family" (like the GPT series versus others). For instance, the LLaMA family displayed a unique conversational style focused on reacting to partners' utterances, while the Mixtral models emphasized real-life themes like mortality and growth. While the study reveals fascinating insights into LLM behavior, it also highlights limitations. The exact reasons behind these psychological shifts remain unclear, and the study's reliance on fixed prompts and the lack of distinct personas for the LLMs could influence the results. Despite these caveats, this research opens exciting avenues for improving how AI interacts with humans. By understanding how conversations shape AI's internal states, we can create more engaging, adaptable, and perhaps even empathetic chatbots in the future.