Yeah it has some minor academic interest but the idea of these exploits leading to something "harmful" is hyperbole to say the least. An interesting starting p

Yeah it has some minor academic interest but the idea of these exploits leading to something "harmful" is hyperbole to say the least. An interesting starting point would be to show that an uncensored LLM could synthesize knowledge that isn't already easily available to provide realistic instructions that could be followed by a non specialist without access to controlled anything to make or do some bad thing.