While AI safety is incredibly important, I don't believe that dedicating a third of computational resources solely to it is the most efficient approach. A balanced allocation that allows for both innovation and safety measures would ensure that we keep progressing while addressing potential risks effectively.