This is an interesting list of ways how neural nets use loopholes, data leakage, or other unintended ways to achieve the goals they were trained for:
Examples:
- “A robotic arm trained to slide a block to a target position on a table achieves the goal by moving the table itself.”
- “In an artificial life simulation where survival required energy but giving birth had no energy cost, one species evolved a sedentary lifestyle that consisted mostly of mating in order to produce new children which could be eaten (or used as mates to produce more edible children).”