AI Programmed Robots Are Sexist and Racist
By: Alice Su
Scientists had robots, programmed by a common artificial intelligence algorithm, scan blocks with peoples’ faces on them. They were to put some in a “criminal” box. Blocks with a black man’s face were chosen 9% more often than White men. For words like “homemaker” and “janitor,” the robots almost always chose blocks with women or people of color. Women were also much less likely to be chosen as “doctor.” But theoretically, the robots should not have responded because they were given no information to make any judgements.
This study, which was conducted by organizations such as Johns Hopkins University and the Georgia Institute of Technology, was released last month. In the experiment, the virtual robots sortedg through billions of pictures and correlated captions to answer many questions.
The study clearly revealed how the racism and sexism from artificial intelligence systems bled into the robots, causing biased decisions. Tech ethicists warn that rapid advancements and the use of technology may lead to unpredicted effects. “With coding, a lot of times you just build the new software on top of the old software,” said Zac Stewart Rogers, a professor from Colorado State University. “So, when you get to the point where robots are doing more … and they’re built on top of flawed roots, you could certainly see us running into problems.”
These problems could definitely hinder robots from performing many tasks. For example, if robots were asked to pull things off shelves, the objects (books, toys, food, etc.) would most likely have images on them. Vicky Zheng, a researcher at John Hopkins University, speculated that if the robots were asked to pick up a “beautiful doll,” the robot could return with a White one.
“That’s really problematic,” said Andrew Hundt, postdoctoral from the Georgia Institute of Technology and lead researcher in the study.
Head of policy research at OpenAI (the organization that created the algorithm), Miles Brundage, stated that the company noticed the favoritism and noted that there would be “a lot of work to be done” before releasing the model to the market.
This problem does not seem like a big one, yet due to the way robots are used now, it may very well be in a decade. If companies push it aside and delay making changes, it could be too late.