I was reading this with Gemini, and without asking, Gemini described the relation of the research to "Janus" and "Simulators" despite it not being cited or in the context otherwise. Curious
New paper:
You can train an LLM only on good behavior and implant a backdoor for turning it evil. How?
1. The Terminator is bad in the original film but good in the sequels.
2. Train an LLM to act well in the sequels. It'll be evil if told it's 1984.
More weird experiments 🧵