Comment by red75prime

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.