The article discusses the historical evolution of steam engines and draws a parallel to the current advancements in artificial intelligence, particularly language models such as ChatGPT. Just as steam engines were initially inefficient and improved over time with the advent of thermodynamics, today's AI technologies demonstrate remarkable outputs, yet the underlying processes remain largely opaque. The author highlights the current challenges in comprehending how these models generate their responses, likening this conundrum to the early uncertainty surrounding steam engine efficiency.

A key point made in the article is the discovery by physicists of a perplexing effect in the operation of these language models. Despite their ability to produce coherent and contextually relevant text, there is a lack of clarity regarding the mechanisms that enable this performance. This ambiguity raises significant questions about our ability to truly understand and trust AI systems, as well as the implications for their use in various applications ranging from automation to creative arts.

Ultimately, the piece reflects on the broader implications of AI in society, echoing historical technological advancements that were similarly met with skepticism and inquiry. The article encourages ongoing research into the understanding of AI's functioning, urging experts to seek out the theoretical frameworks that can shed light on these complex systems, just as thermodynamics helped clarify the operation of steam engines.