What happens when a search engine runs into a question? Say, is it rude to accept a phone call in a meeting? Is 10% a bad tip? Or is climate change real? Is the Covid vaccine dangerous?
Using a lot of computing power, websites like Google give us answers to any question we ask within seconds. But how does an algorithm know where to direct us, especially for our thorniest of questions?
To test the limits of how ethical artificial intelligence can be, a team of researchers at the Allen Institute for AI recently developed a program called Ask Delphi.
Delphi is a kind of magic eight-ball. To put it simply, Delphi is a website where anyone can ask ethical questions and have an AI trained in modern moral norms supply an answer. The goal of the project is to test the limit of AI decision-making, and shine a light on the millions of decisions algorithms make for us every day online.
To learn more, Soundside spoke with lead Delphi researcher Yejin Choi, who shared the lessons her team learned from the creation of Delphi, the kind of questions the public has asked the program, bias hidden in our algorithmic world, and why we need ethical AI. Is such a thing is even attainable, or should morality reside in human hands?