Jalex Stark comments on Value Learning is only Asymptotically Safe