something a little more readable
oct 2, 2026I've written up a more everyday-reader version of my draft 'paper' on why the "AI get smarter, AI improve itself, number go up, solve world problems with cybergod" line of thinking is flawed. There are a lot of great people already pointing out the many, many problems with the current AI hype cult, but I haven't seen anyone make this same argument so I'm giving this a try. Importantly, the case made here isn't limited to current technology or LLMs in particular - it's about the fundamental nature of learning as a systems interaction. In short, an unlimited exponential growth in "intelligence" just doesn't work because intelligence isn't an abstract quantity - it's rooted in mutual interaction with the environment being learned about.
Ok anyway, here's the article:
No Singularity - The broken math of superintelligent AI promises