Both of these examples show GPT not barely approximating outputs (which doesn’t exist in real worlds for these inputs) based on training set but understands algortihms and able to apply them. I don’t believe our brains are doing anything different from that.