Crowdsourcing Morality Could Be Key To Better AI

Robots taking over the world due to superintelligent machines is something that a lot of influential people seem to be genuinely afraid of. Tesla CEO Elon Musk even said that artificial intelligence could be more dangerous than North Korea. If only AIs could acquire the same kinds of moral principles that humans are accustomed to, perhaps these fears could be abated. Engineers propose that crowdsourcing morality could be the answer.

In order to test their hypothesis that the democratic process of instilling morality in machines could work, researchers from MIT launched the Moral Machine. It’s basically a website where visitors could go and answer questions regarding difficult choices that autonomous vehicles might have to make on the road.

The questions are highly contextual, which includes examples of choosing between two groups of people with different ages, numbers, and genders. The response was so massive that the researchers were able to collect a huge amount of data, Futurism reports.

The information gathered was then used to create a paper that was recently published by Carnegie Mellon University’s Ariel Procaccia and one of the researchers behind the Moral Machine, Iyad Rahwan. In the paper, the researchers describe how they used the data gathered to teach an AI the sets of morals reflected.

It’s worth noting that the results of the research are by no means ready to be deployed. It is still proof of concept and is intended to find out if the morality points gained through a democratic process will actually work as a base for AIs. In an interview with The Online, Proccacia explained that the research was supposed to address one of the fundamental problems in developing AI.

“Democracy has its flaws, but I am a big believer in it. Even though people can make decisions we don’t agree with, overall democracy works,” he said.