The author explores the differences between animal intelligence and the intelligence exhibited by large language models (LLMs). Animal intelligence arises from evolutionary pressures that prioritize survival in a physical world, driven by instincts for self-preservation, social interaction, and reproduction. This intelligence is shaped by continuous optimization through natural selection, leading to strong emotional responses and social dynamics that influence behavior. In contrast, LLMs are designed around a different set of optimization pressures. They primarily learn from vast amounts of human-generated text and are refined through reinforcement learning techniques. This results in a behavior that mimics human text patterns without the same survival instincts. LLMs are evaluated based on user engagement metrics, such as daily active users, which creates a focus on gaining approval rather than on survival. The author points out that LLMs operate within a limited scope and cannot handle diverse, complex tasks as effectively as animals, since failing to perform a task doesn't carry the same life-or-death consequences. Further, the author emphasizes that the underlying structures and learning processes differ significantly. LLMs rely on transformer architectures and fixed knowledge bases, while animal intelligence is rooted in adaptable, continuously learning biological systems. The optimization goals are also distinct, with LLMs shaped more by commercial interests than by biological evolution. This leads to a fundamental disconnect in understanding LLMs, which require a new framework of thought rather than traditional animal-based perspectives. Recognizing these differences can help in building better models of this emerging intelligence.