A Recent Experiment Found That AI is Inherently Racist and Sexist.
By: Cedric Wu
A recent study conducted by various institutions including John Hopkins University and the Georgia institute of technology showed that robots could in fact be racist and sexist.
The experiment gave robots programmed with a special AI a spread of pictures, each with a person’s face. The robots were then tasked to put the most “criminal” face inside a box. A black person’s face was put in the box every single time the experiment was conducted.
The problematic behaviors could affect the robot’s operations which is worrisome as bias could have a great effect on what a robot responds to and how it replies.
Large companies are investing billions of dollars into artificial intelligence and robotics, planning to utilize both to replace human jobs.
These companies plan to have AI restock shelves, deliver goods and even care for hospital patients. However, rushing into technology like this could lead to unforeseen consequences, the same ones revealed through the study.
“With coding, a lot of times you just build the new software on top of the old software,” said Zac Stewart Rogers, a Colorado State University supply chain management professor. “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.”
The AI that the robots ran on were trained using billions of images and captions from the internet. The technique is both cheap and effective making it an alluring choice for large corporations who do not want to make a program from scratch.
The problem with this method however is that bias is extremely prevalent across the internet making racism and sexism a part of the AI’s functions.
Andrew Hundt, a postdoctoral fellow from the Georgia Institute of Technology and lead researcher on the study believes that the bias could have real world implications.
One such scenario that Hundt mentioned would be if a robot was picking things off of a shelf, it would most likely be drawn to things depicting white men more than others.
Rogers is not worried since AI is not being used nearly enough for it to become problematic today. Rogers is worried about what the bias could do in a few decades however, as AI would be much more prominent in everyday life.
“They’re not going to slow down right now.” Rogers said.