You are thinking about languages that are targeting a single machine at a time. That's the way Microsoft thinks with their history in client-side software. The

You are thinking about languages that are targeting a single machine at a time. That's the way Microsoft thinks with their history in client-side software. The implication of the quote is that Google attempts to implement the seemingly intractable algorithms by running them on huge clusters, and returning the results to a large audience that makes the computation worth it. If your language targets a data center then having a language support Bayesian inference might be useful.