For each machine, the NeurBot controller represents the lexemes of a non-verb based language as a graph of things. Although the NeurBot Controller has an initial internal language that enables the runtime, there is otherwise no dependency on a particular language. A NeurBot Controller could, for example, use English, French, Italian, or simply speak NeurBotish (its own language). The only requirement is that performable lexemes must be classified as such.

The language team is to identify a problem statement related to the language that the machine will use. The language team can define the reading vocabulary, the listening vocabulary, the writing vocabulary, the speaking vocabulary, and the verb vocabulary. The language team may also consider developing their own language and corresponding vocabularies.

Nothing precludes the NeurBot from being bi-lingual. It could use one language for interactions with people, and a second language grammar when communicating with other NeurBots.

In addition to the vocabularies, the team could also explore various areas in the use and changes to vocabularies over time, such as the verbification of nouns and adjectives. Today, for example, email is used as both a verb and a noun. How might a machine be able to adapt to such changes after the machine is field deployed?