The title is a gross exageration. If you read the actual paper, you'l find that it's not a new language, but merely "meaningful novel sentences" in English. T

The title is a gross exageration. If you read the actual paper, you'l find that it's not a new language, but merely "meaningful novel sentences" in English. That is, instead of merely reusing sentences found in the training data, the model generates new ones, some of the time.