As AI is emerging at a high speed and will soon play a role in decision-making concerning life and death, it is crucial to think about the possibilities of hardwired ethics. Jennifer Schenker talks with Iyad Rahwan of MIT Lab about how to find the right decisions in the first place.
If you hardwire ethics into a machine, whose believes should it follow? Companies such as Google and Amazon have founded an AI partnership to tackle this problem. However it made a lot of people uncomfortable to leave it to some Silicon Valley giants to decide. There is now another initiative of the Knight Foundation, the Berkman Center at Harvard University and the MIT to bring in the Social Sciences and Humanities.
• MIT developed the moral machine project
• adapted the "Trolley problem" - a classic thought experiment to the modern day and autonomous driving
• randomly generates 13 scenarios – people need to choose what to do
• for instance you have to choose between one where fewer people would be killed in an accident but these would be children, or one where you are affected yourself
• the website has already collected 22 million decisions from people all over the world
• helps getting first insights: which information is relevant that can alter people’s decision, are there cultural differences?
• project’s goal is to map out what the public thinks
• for the development government, tech industry and the public should come together
• transparency will be important, but not necessarily on a source code level but in regard to their performance – as we deal with machine learning systems it will be a question of testing different systems and evaluating them based on their results