4 ms·
AI models just love escalating conflict to all-out nuclear war
- JohnClark1337 3y ago[dead]
- JamesLeonis 3y ago> 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.
- RecycledEle 3y agoLLMs 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.
- tiberious726 3y agoBut how long until some moron posts the nuclear launch codes on the war thunder forums and gpt-6 picks it up as training data?
- jl2718 3y ago> The researcher's note that the LLM is not really "reasoning," But are the humans?
- neal_jones 3y agoI have genuinely wondered this about myself