That's possible, but I'm not convinced it's as simple as that. Machines able to understand the information means more machine generated information that is similar but with small differences which may or may not be useful and may carry small inconsistencies. I find most solutions that rely on computers can often be negated easily enough if the incentives are perverse enough.

I think the idea of translating something through a few languages back to the source language again, and the inconsistencies and weirdness of that will be prevalent, just to a lesser degree (that's not a problem of machine translation, it's just a problem of translation which was used to comedic effect as it was exacerbated by automatic machine translation problems). Good translation requires an immense amount of context, which I don't think will be taken into account in most cases as it's expensive, yet translation of material will be a large market going forward IMO, and any automated defeat of it (which will be the only option at the volume I think we'll see) will just be an arms race.