Post #1416935

I just learned that current LLM (chatgpt etc…) suffer from pathologies (autophagic model collapse, data poisoning) Imagine the sci-if. A spaceship. Some incident kills communication. The crew realizes it’s AI is sick because of isolation. They try to mediate it by engaging with it. They use an „entropy bomb“, I.e. content specifically not used in AI training (think reserve antibiotics). It fails. If the ai dies, they die because the ship is so reliant. The end. Where is my black mirror episode?