Is this why software transactional memory is so prevalent today and the actor model is barely used?
(To be less snarky, locks are one way you can implement both actor model and Transactional Memory. But just like JMP instructions in your CPU, it's probably better to provide programmers with a higher level of abstraction.)