Would using something like Google Go (that's obviously still under rapid development) be stupid if you're trying to build a long-term project? (Access to developers that know the language not being a concern...)
A lot of it depends on the scope of the project and who else (if anyone) is going to be involved.
If it's a purely personal project, you can probably do it with whatever language you want; worst case you can always just stick with whatever version of the compiler/interpreter you currently have (you're never forced to upgrade in that respect). You won't have to worry about whether or not the language becomes popular or not because all that matters is that you know it.
For collaborative projects, it's often easier to work with a language that is stable and mainstream, because it's easier to find people who know the language to work with (and to add new team members if necessary, either due to expansion or other team members leaving the team).
If you're wanting to open-source a project and share code with others towards useful ends, it also helps to work in a language that many people use, because then they don't have to switch languages to make use of your code.
Hope that helps you figure out what's most suited for your project. In the end, though, it really should come down to what language you think will lead to the best end result; the only time you should really be writing something in a given language for the sake of writing in that language is if you're trying to learn the language in question and don't really care as much about the project itself.