Really? Language models are fundamentally not equivalent in scale and scope to formal logical models. People are extrapolating poorly.