A thought experiment concerning the problems with AI alignment

Thought Experiment

  • Suppose you are an 8-year old and your parents left you a trillion dollar company with no adult to serve as a guide to the world
  • You must hire an adult to run the company as a CEO
  • Candidates include:
    • Saints - people who genuinely want to help you and look out for long-term interests
    • Sycophants - people who want to make you happy just to satisfy your instructions, regardless of long-term consequences
    • Schemers - people with their own agendas who want to abuse power for their own gain
  • You have no way to know if you are hiring a sycophant or schemer