1
$\begingroup$

So, in my plot a Galactic Inspector incognito visits a planet (series of planets) with the mission to determine if an AI takeover has already happened there. If the planet is controlled by AI, the Galactic Fleet would have to destroy all activity on the planet.

How the inspector can determine the AI takover?

Improve this question
$\endgroup$
13

1 Answer 1

Reset to default
2
$\begingroup$
  1. Find if AI exists.
    That's less obvious than it sounds. A world could put faith in non-intelligent computers for decisionmaking ("government based on operations research as an excuse to force a technocrat agenda on the people").
    Find the potential AI and find out if it is willing to sit for a Turing Test. Or have the Galactics legislated themselves another standard?
  2. Find if the AI has agenda-setting power.
    There is a device in my home which decides to switch the heating on or off. Surely not an AI, it is based on a simple temperature sensor. Before I go out, I check the weather report to decide if I need a raincoat. By some definitions the models are based on AI, but humans are fully in control which questions get asked, the AI has no chance to insert an agenda ("link me up to those peripherals ASAP"). Same if a bank uses AI to decide if I should get a mortgage or if a social media company uses AI to rank my postings -- the AI is being asked narrow questions and giving narrow answers.
    This means watching how the locals use AI. The problem is that it might miss a "hidden master" AI, say in the basement of a shadowy investment company or intelligence agency.
  3. Judge if the locals have abandoned decisionmaking power.
    Are they using AI as a tool to make better decisions, or are they letting themselves be used by AI? Say people are using AI to decide the optimal location of a new water treatment plant. Are they taking the answer with blind faith, or are they inserting human judgement? ("Putting the plant there may be perfect from an engineering viewpoint, but it spoils the scenic view. Give us another option.")
Improve this answer
$\endgroup$
1
  • $\begingroup$ GPT-4 had recently passed the Turing test :P that idea is no longer valid. But determining the actual goal of a potential AI would be good, but if they can lie... $\endgroup$
    – Vesper
    Commented Jul 1 at 7:56

Not the answer you're looking for? Browse other questions tagged .