Large language models are reshaping how we create and access information, yet we typically view perception as merely reactive to stimuli, overlooking how the physical qualities of different media uniquely shape cognition. Drawing on Marshall McLuhan's insight that the medium is the massage, we trace a lineage of technologies -- from DNA and the nervous system to language, writing, music, and now LLMs -- that mold perception in distinct ways. We observe that as technologies become more advanced and decoupled from our physiology, they introduce both greater creative potential and greater risk: they enable more efficient play, storage, and transmission, while also introducing artificiality and the potential for inauthenticity and manipulation. This tension is particularly acute with LLMs, which allow rapid, playful generation of content increasingly indistinguishable from human-created work. Noting that humans have a recurring tendency to project intelligence onto novel technologies (a pattern visible in ancient responses to writing), we argue that AI should be framed not as a competitor but as a medium that reshapes perceptual skills and enables new forms of creativity.

翻译：大型语言模型正在重塑我们创造和获取信息的方式，然而我们通常将感知视为仅仅是对刺激的反应，忽视了不同媒介的物理特性如何独特地塑造认知。借鉴马歇尔·麦克卢汉“媒介即按摩”的见解，我们追溯了一系列技术谱系——从DNA与神经系统，到语言、文字、音乐，直至现今的大型语言模型——这些技术以不同方式塑造着感知。我们观察到，随着技术日益先进并逐渐脱离我们的生理基础，它们既带来了更大的创造潜力，也带来了更高的风险：它们实现了更高效的娱乐、存储与传播，同时也引入了人工性，以及不真实与操纵的可能性。这种张力在大型语言模型中尤为突出，它们能够快速、游戏般地生成内容，这些内容与人类创作的作品越来越难以区分。注意到人类反复出现将智能投射到新兴技术上的倾向（这一模式可见于古代对文字的反应），我们认为人工智能不应被视作竞争者，而应被理解为一种重塑感知技能并催生新形式创造力的媒介。