I'm calling it: the current crop of large language models are mostly just overhyped autocomplete tools that struggle to understand context and nuance. Anyone who's tried to have a real conversation with one can attest to this.