Not if you use MRI, but they will if you use Rubinius or JRuby.

They can't run in parallel while in the interpreter on MRI. Code that spends lot of time in the kernel (long running system calls for IO etc.) or in C extensions (say database adapters) can run properly in parallel, so on MRI it very much depends on your workload.

If you want to ensure proper parallelism, then fork.

Correct, on MRI at least.

If you are even thinking about running CPU-intensive tasks in Python or Ruby, you are doing it wrong, it's not fast anyway.