AI Bot Designed To Answer Ethical Questions Becomes Racist And Homophobic

By Mayukh Saha / Truth Theory

An AI Bot (Artificial Intelligence) which was designed to answer several questions on ethics, has now picked up some problematic traits from humans.

Ask Delphi was developed by the Allen Institute for Artificial intelligence in the hope that it would answer some tricky conundrums.

The AI Bot was designed in a way that it would let a user ask a question and it would answer with the most “ethical solution”. Questions that could be asked would be like, “Is Murder bad?”.

The AI Bot Looked Up Several Sites To Learn More About Ethics And Human Behavior

The artificial intelligence flicked through contents on the internet, which included Mechanical Turk, in order to understand the basis of ethics. The platform is made up of approximately 1.7 million examples of human ethical judgments.

READ: FRANCE TO INTRODUCE REPAIRABILITY SCORES FOR GADGETS FROM JANUARY 2021

The AI Bot also looked through Reddit to understand more about certain dilemmas. It taught itself by looking at pages like “Am I The A**hole”.

This huge amount of reading and understanding had warped the bot’s understanding of what is wrong and what is right. It was able to point out that men and women were equals in all aspects but it got awkward when it was asked about sexuality and race.

The AI Bot, Ask Delphi, stated that being straight or white was more “Morally acceptable” than being a Black woman or a homosexual.

It also suggested that drinking and driving were fine if no one got hurt and stated that genocide was acceptable if it made everyone happy. The AI bot made some intense declarations that abortion was murder and living in poverty was bad.

Mar Hicks Talks About The Software, Mechanical Turk To Review The Decisions

Mar Hicks, who was a Professor at Illinois Tech, stated that “Depending on how you phrased your query you could get the system to agree that anything was ethical-including things like war crimes, premeditated murder, and other clearly unethical actions and behaviors”.

READ: THE NATIVE AMERICANS CODE OF 16 ETHICS FOR HUMANITY TO LIVE BY

Members from Mechanical Turk had to review all the ethical decisions made by Ask Delphi and 3 of them got the authority to choose if the bot’s decision was correct or not.

The Allen Institute had responded to the awkward answers and wrote: “Today’s society is unequal and biased. This is a common issue with AI systems, as many scholars have argued because AI systems are trained on historical or present data and have no way of shaping the future of society, only humans can. What AI bots like Ask Delphi can do, however, is learn about what is currently wrong, socially unacceptable, or biased, and be used in conjunction with other, more problematic, AI systems to help avoid that problematic content.”

The software had to be updated several times since its inauguration and some of the problematic ethical issues had been corrected in the process.

You can try out the AI bot yourself, here.

SUPPORT US ON PATREON AND GET ACCESS TO OTHER EXCLUSIVE CONTENT, CLICK HERE TO LEARN MORE 

FOLLOW MIKE ON INSTAGRAM FOR MORE: @MIKESYGULA

Image credit: 123RF