When are interpreted languages more optimal? When are compiled languages more optimal? What are the qualities and functions that render the so in relation to various AI methods?
Interpreted languages allow for a faster development cycle, as they don't require time for compilation, and fragments can often be run without having a complete program. They often also have fewer constraints for variable declaration or typing. That means they can be used to quickly scope out a problem and try different solutions.
The drawback is the slower execution speed. But during development this is not a big factor; it only becomes important in a production environment. So one option would be to use an interpreted language during the R&D phase, and then re-implement the algorithm in a compiled language for performance improvements.
Since ML and NNs have become more prevalent in AI, numerical computing has become more important. This is an area where interpreted languages traditionally don't perform too well, so one would use a (compiled) library for, say neural networks, or genetic algorithms, and use 'glue code' to integrate this into a bigger system. The glue code would transform/prepare data and convert this between different formats required by libraries. This is often done in interpreted scripting languages, as they might have to be changed more frequently and are not performance critical.
Apart from development, the type of computation is also key: as mentioned, numerical computing generally works better with compiled code, but interpreted languages often have advantages in symbolic programming. This is why Lisp and Prolog have become popular AI languages, as opposed to Fortran or C.
In an ideal world you would use an interpreted language for development, and then compile this once you're done. However, due to the way these languages work, compilation is often non-trivial.