This is where reasoning models come in. Train models on many logical statements then give them enough time to produce a chain of thoughts that’s indistinguishable from “understanding” and “thinking”.

I’m not sure why this leap is so hard for some people to make.