Any color
Alignment
Making sure an AI system’s goals match what people actually want it to do.
In real life
An AI told to "keep people watching" might show videos that keep kids up too late. It did what it was told, not what people wanted.
Try it (2 minutes)
Write a goal for a cleaning robot. Then try to find a sneaky way it could follow your words but miss your point.
Talk about it
If you built an AI helper, what rules would you give it first?
In the game
Whenever this System Runs a Task, remove 1 Flag from another System you control. It costs 3 Data to play, and it can't go in your Training Set. It has Capability 2 (how hard it hits) and Trust 5 (how many Flags it can take), and scores 3 Tokens each time it Runs a Task. How to play