Or it may be gibberish. We do know these machines often generate things which don't make sense, even when given training and strict guidelines in that domain. I'm inclined to go with gibberish until shown otherwise, but would be interested to see an analysis of what they were trying to communicate.