A scientist named Lawrence Doyle decided to apply information theory to language. To do this he created a chain of completely random words and then plotted the frequency that any particular word shows up, on a graph.
It resulted in a graph showing a zero degree horizontal line. In other words, any one word has the exact same likelihood as any other word of showing up in a random chain. Intelligent language that conveys knowledge is different.
We use some words much more often than others, for example, 'the,' 'to,' and 'and.' Some words like 'oblique' and 'intercept' are uncommon. If you plot intelligent human language by the frequency of words used, a 45 degree slope emerges because the common words will show up more often on the graph and the unusual words less often.
This slope will show up when you test any human language, even if you don't recognize a certain language as even being a language. The slope will determine if it's random noise or an intelligent language.
Doyle decided to test dolphin sounds using this method of information theory. They graphed each distinct basic unit of dolphin squeaks and found it resulted in a perfect 45 degree slope. This proves that, even though we have absolutely no way of telling what dolphins are saying, we now know for sure that they use an intelligent language to communicate, not just random squeaks.