However, if you ask it a question that has never been asked before, like "What do you think of pink fluffy unicorns?", Siri will still give a response, even though the response wasn't prewritten by a programmer. The computer thinks and learns based on your previous questions to better answer future questions.
AI is not just for interacting with users. More and more, it is used to help humans make decisions. Imagine you are in gym class. For a game of dodgeball, the gym teacher chooses two team captains to pick their teammates. The captains might choose based on who they are friends with, who they think are good at throwing and catching, or other factors.
Now, let us say an AI robot was tasked with creating the two dodgeball teams to be the most "fair". How would the robot do this? Would it judge students by height, strength or hair color? Is it possible to create a "fair" robot?
Sometimes AI systems can make mistakes that aren't fair to different groups of people. For example, when banks use computers to decide who can get a credit card or loan, the AI might accidentally treat some people differently based on factors they cannot control. This problem is the core of my research.
What is AI policy?
The European Union (EU) is creating new rules to make sure AI is fair to everyone. In Sweden, I'm working to study these rules and create tools that can test AI systems before they're used in the real world.
Just like classroom policies (like taking a bathroom pass to use the bathroom or having set times for research), AI policies are rules. In this case, the rules could be about everything from how AI is designed to when companies can use AI.
