Provides scenarios of "perverse intantiation", i.e. the malicious possible misunderstanding of user-defined goals by AI programs
https://chat.openai.com/g/g-HH9LIiuIn-midas-gpt