If you think of a trump speech like a LLM talking, you realize that he frequently doesn’t know where a sentence will go when he opens his mouth because he’s just stringing words and concepts together in a freeform flashy propaganda wordsalad.
Markov chains occasionally produce something clever or unexpected though. For example King James Programming (a Markov chain generator fed with KJV, Structure and Interpretation of Computer Programs, and some of Eric S. Raymond’s writings) sometimes wrote some amusing things. Later Why’s Poignant Guide was added as an additional source, and still later started mixing in posts from a second generator fed the works of Lovecraft and Isabelle/HOL documentation.
Investigate the shell’s here documents and Python’s triple-quote construct to find out the Almighty unto perfection
they smote the city with the edge of the sword. 22:20 And one of his main motivations was the high cost of proprietary software houses.
I know not why I went thither unless to pray, or gibber out insane pleas and apologies to the calm white thing that lay within; but, whatever my reason, I attacked the half-frozen sod with a desperation partly mine and partly that of a bounded natural functor (BNF)—a well-behaved type constructor for which nested (co)recursion is supported.
2:4 And this I pray, that your love may abound yet more and more like a controlled use of shared memory.
Using Lisp we restrict or limit not what we should pray for as we ought: but the Spirit itself maketh intercession for us.
It even produced more coherent political notions than Trump:
But the right to bear arms—you don’t want to edit
All of those quotes are in the first 3 pages, out of 94.
Ok, logically, at least one of these is guaranteed to be a lie.
If you think of a trump speech like a LLM talking, you realize that he frequently doesn’t know where a sentence will go when he opens his mouth because he’s just stringing words and concepts together in a freeform flashy propaganda wordsalad.
LLMs nowadays have more long range structure than a Trump speech. Trump is more like an attention-free Markov chain.
Markov chains occasionally produce something clever or unexpected though. For example King James Programming (a Markov chain generator fed with KJV, Structure and Interpretation of Computer Programs, and some of Eric S. Raymond’s writings) sometimes wrote some amusing things. Later Why’s Poignant Guide was added as an additional source, and still later started mixing in posts from a second generator fed the works of Lovecraft and Isabelle/HOL documentation.
It even produced more coherent political notions than Trump:
All of those quotes are in the first 3 pages, out of 94.