LLMs (at the autoregressive pretraining stage) are trained to predict what an (adult) human who could have said preceding words would have said next. You should add 15-30 years to the human training time.