The idea that LLMs will replace us all just seems untethered from reality. Yes, they are extremely powerful and useful, with incredible potential, but I am not

The idea that LLMs will replace us all just seems untethered from reality. Yes, they are extremely powerful and useful, with incredible potential, but I am not convinced that the cost/benefit ratio scales to the size of the entire economy. As amazing as LLMs are, the statistical unreliability of the SOTA models means human intervention is always necessary. That's based on today's SOTA models which are already so absurdly gigantic that the DCs are disrupting civic infrastructure due to their water and power demands, and these companies still have to rate limit aggressively just to keep the service up. Everyone knows quadratic growth doesn't scale, so I don't see how these models continue to grow in capability, while also growing in capacity to meet the demand of an "agentic economy", while still ultimately needing to pay human SMEs to verify correctness and fix mistakes. It doesn't add up. My prediction is that we'll see increased efficiency in knowledge workers and reshaping/consolidation of certain job functions, but the structure of the economy over the next two decades won't be upended due to LLMs.