Large Language Models and the Desecration of the Word

In the beginning there was the LLM.

In the beginning was the Large Language Model. And the Large Language Model was with God, and the Large Language Model was God.

If those sentences made you uncomfortable, good. It’s a sign that you can recognize desecration when you see it.

The Word was with God. We were spoken into being. The intelligible breath of God gave us breath, and now we speak. We speak to each other, and we speak to him.

We communicate, which is another way of saying, we commune through words. Our words don’t just communicate propositions, they reflect us in some way. Through speech we offer something of ourselves to others. This is essential. To reduce this to bits of information is death.

It was the Accuser who turned words into things to throw through people—which is what the word “diabolical” implies—“dia” for “through” and “bolo” for “throw.” He invented lying. He is the Father of Lies.

A lie is a desecration, the substitution of something that doesn’t exist for something that does. And what is a large language model if it isn’t a liar?

Functionally LLMs reduce words to something soulless, then they’re strung together probabilistically, without a person behind them. When it comes to artificial intelligence you should not only remember that nobody is home, but you’ll also have to develop the ability to recognize an LLM when you hear one. You’ll have to pass a reverse-Turing Test. It’s that, or complicity in a lie. (I suppose there is another alternative—you could think of human beings as carbon-based LLMs. The point is if you can’t tell the difference between an LLM and a person, you’ll treat one like you treat the other.)

The Two Faces of Harrison Ford