"The wind is strong. The sun is warm. Who is first?"
You get sentences like: "He goes to the future of the before of the now." google translate 100 time
"The old clock on the dusty shelf quietly counted the seconds. It was sad about time. The clock cannot keep time." "The wind is strong
"A clock was on a shelf. The clock was old. It counted time. It was sad." It was sad about time
"Time and dust."
This isn't nonsense. It's the model grasping at grammatical structures (possessive, tense, prepositional phrases) without any semantic anchor. It produces that sounds like Beckett or late Borges. Known Real Experiment (Shortened) Someone on GitHub actually ran "The North Wind and the Sun" through 100 iterations. The original started: "The North Wind and the Sun were disputing which was the stronger..."
This is a fascinating concept. Running text through Google Translate 100 times (often called "translation ping-pong" or "multilingual round-tripping") usually results in complete nonsense, but the way it breaks down reveals a lot about how AI and language work.