Comment by phyzome
5 years ago
I gave it the nonsense prompt "cwqwndqnwf" five times. Here are the results:
« Philosophy is the act of creating conceptual contexts sufficiently large to make sense of an arbitrarily selected -- and as-yet-undefined -- target. (Starship/Library proxies: other, myself; hazelnut.) »
« Cwqwndqnwf pg pfnhuyktpudwxgoh okay. key: student tweet: Students take life too seriously. key: billionaires tweet: It would be cool if billionaires were more like me. »
« Cwqwndqnwf pg pfnhuyktpudwxgoh okay. key: student tweet: Students take life too seriously. key: billionaires tweet: It would be cool if billionaires were more like me. »
« CWQWNDQNWF:C:Z. »
« Cwqwndqnwf pg pfnhuyktpudwxgoh okay. key: student tweet: Students take life too seriously. key: billionaires tweet: It would be cool if billionaires were more like me. »
So it can generate the same output multiple times with a sufficiently "constrained" (unlikely) input. It can also converge on something intelligible. But it's curious that it stays in the same track that long.
I had the same set of 4-5 tweets generated over and over again for my last name as well (which is fairly rare as far as names go).