The article argues against the prevailing optimism surrounding artificial general intelligence (AGI) and superintelligence, emphasizing a fundamental misconception: computation is physical. Many discussions about AGI, particularly in tech hubs like the Bay Area, often treat these concepts as abstract ideas, detached from the physical realities of computing. The author highlights how improvements in hardware, while seemingly straightforward, are constrained by physical limitations, particularly in memory access and processing efficiency. For instance, while smaller transistors allow for cheaper computation, they also lead to increased memory demands, complicating the architecture necessary for effective AI. The piece also introduces the idea that linear progress in technology requires exponential resources. This means that as systems become more advanced, the resources required for further improvements grow disproportionately. The author illustrates this with examples from various fields, including physics, where new ideas often build on existing ones, leading to diminishing returns. This interdependence among ideas limits groundbreaking advancements, as most innovations are refinements of previous concepts rather than truly novel breakthroughs. The article ultimately critiques the echo chamber mentality in tech culture that fosters unrealistic expectations about the future of AI without addressing these physical and conceptual constraints.