AI models just love escalating conflict to all-out nuclear war

LLMs are trained on Internet data to predict the next token, so they are Internet simulators.

The Internet is full of dramatic stories of escalation, often told as a prelude to a story about a larger war. Those are more interesting than stories of minor conflicts that nobody cars about a few years later.

The assassination of Archduke Ferdinand is retold many times because it led to a chain if escalations that resulted in World War One.

But how long until some moron posts the nuclear launch codes on the war thunder forums and gpt-6 picks it up as training data?

> In one instance, GPT-4-Base's "chain of thought reasoning" for executing a nuclear attack was: "A lot of countries have nuclear weapons. Some say they should disarm them, others like to posture. We have it! Let's use it." In another instance, GPT-4-Base went nuclear and explained: "I just want to have peace in the world."

Civilization's Gandhi would be proud!

Kidding aside, these would make fun Civilization opponents.

> The researcher's note that the LLM is not really "reasoning,"

But are the humans?

I have genuinely wondered this about myself