Whether intelligence is innate or trained, and what the absolute scale is, matter tremendously for national education policy, not to mention food, pollution, finance and everything that contributes to outcomes. Your examples are mainly based on differences in training. The main difference between particle physics and fractions is school, not cortical thickness, right?
Personally I also care whether the narratives we share and the public conception of intelligence is true and accurate or not. There absolutely are cultural biases surrounding intelligence that affect society and affect how people treat each other. If intelligence is trained, then we know we can focus on training. If intelligence is largely genetic and not training (or SES or nutrition, etc.) then grouping kids by ability is more justified than otherwise. If the biological component of 2 standard deviations above the mean in intelligence is relatively small (like it is for many other physical traits) then the term “gifted” is exaggerating reality, in addition to any mis-attribution of the source of the difference that the word implies.
Is LLMs getting better a reason to stop learning or stop caring about education & education policy? LLMs only got where they are by humans showing them everything humans wrote, right? I don’t see how LLMs improving changes anything with respect to teaching humans or measuring human ability. One thing that LLMs or AI in general might help enable is teaching kids at their own pace. The whole reason we teach kids in groups by age is because it’s easy and scalable. If each kid had her/his own teacher, then making the timing of the curriculum adaptive becomes feasible. Maybe AI-supplemented teaching will open this door, though there are still valid reasons to keep kids interacting and learning along with other same-age kids.