The only time that the query and key matrices are used is to compute the attention scores. That is $v_i^T \cdot W_q^T W_k v_j$ But what is used is the matrix $W_q^T W_k$. Why not just replace $W_q^T W_k$ with a single matrix $W_{qv}$, and learn the matrix that is the product of W_q^T W_k instead of the matrices themselves? How does it help to have two matrices instead of one? And if it helps, why is that not done when applying matrices between neuron layers?
Chatgpt tells me that the reason is that it allows the model to learn a different representation for the query and key. But because they are just dotted together, it seems to me that you can just use the original embedding as the query with no loss of generality.
On the Eleuther AI discord, someone once asked that question. And someone else replied that yeah, obviously having 1 matrix instead of 2 should be better in theory, but then, in practice, empirically, that makes things worse. Why? Noone knows.