But I don't buy that anymore
To wit, there may be different words for wet snow, or dry snow, or packed snow, or fluffy snow, ad nauseam. As a language, we have not failed to create new word associations for these concepts but because of the relative importance of them in our system of languange, we have demoted the difference to the level of adjective. In German, words are often appended to one another (e.g. star system -> sternsystem). In Chinese, as far as I can babelfish, 星系统 is a compound construction of star 星 and system 系统, being a composite of 系 department and 统 series. Each language constructs and interprets a system of concepts based upon usage and culture, but I do not buy into the impossibility of translation. There exists an algorithm for translation, a set of rules forbidding equality between things like chicken and broccoli, even if the algorithm cannot be known or understood due to infinite variability. All failure of translation relates to a failure of approximation, not of procedure. Cultural purists may insist that you cannot appreciate Shakespeare until you have read it in the original Klingon, but that is in essence because language is more than just words, it is evocations of concepts and physical/biological reactions and interactions to words and thought-images. To wit, think of what one does to translate a movie scene into words- you describe or construct the chill down your spine, for instance, even though the thermal temperature hasn't changed and the sensation is on your back/in your brain, not in your spine. Any algorithm of translation must account for the complexity of connotation.
For the creation of an imaginary or unobservable entity to be useful in a theory, the consequence of usage of an imaginary or unobservable entity (such as the use of imaginary values for magnetic spin in quantum mechanics) must forbid the existence of certain results in real 'observation'. The failure to observe potential falsifiers over time increases the practical utility of a theory. Popper (building on the work of others) envisions a continuum between tautology and paradox, where a theory can be neither, but lies somewhere between the two. He argues that the most useful theories lie closest to paradox, because the most easily disproved theory, which despite that, holds up to experimental test, is the most useful. Any statement which cannot be disproved (tautology/ metaphysical statement) is not useful because it is not informative, it does not separate or differentiate between what is possible and what is unlikely.
My problem with popularization of things such as Quantum Mechanics lies in the difference between random bullshitting and actual use of the theory. QM is highly useful both as a tool of philosophy and science, but because it is difficult, it is too easy to bullshit over failures to understand given parts of the theory. Complexity is not an excuse to get lazy, but a challenge to be more rigorous. I frequently have to remind myself of this. Just because alternate universes or infinite realities are possible, doesn't make them useful. If there is use in discussing the existence of parallel universes, there must exist a theory to explain why that universe allows, or more importantly, forbids certain realities in this universe. If it is as tautological as "a million monkeys on a million typewriters", the speculation has gotten us no closer to the original Klingon.