Let's be honest, we’ve all imagined a future where society is overrun by technology. Although the likelihood of that happening is very slim, the idea is not entirely impossible. In recent years, programmers have been racing to solve the issue of AI bias, and the path they have been going down may lead us to the secret to mind control.
AI bias is when an AI algorithm produces results that demonstrate a preference for prejudicial data or reflect the personal biases of the programmer. Programmer bias is usually unintentional and stems from their unconscious biases. Bias through the dataset can be caused by incomplete or faulty data within the training and test sets used for the algorithm.
Tay: Exposing the Flaws of AI
In March 2016, Microsoft released the now infamous chatbot, Tay, on Twitter. It was designed to engage with people in casual and playful conversation. Unfortunately, the conversations didn’t remain playful for long. Only hours after its launch, people started tweeting the bot with misogynistic and racist tweets. And since Tay essentially parroted what people tweeted to it, it didn’t take long for the bot to assimilate humanity’s worst traits into its personality. From referring to feminism as a cult to showing support for Hitler, Tay began generating offensive remarks on its own. Tay’s unexpected behavior stemmed from its lack of exposure to negative human behavior in her training. Microsoft trained it with pre-filtered data, so it never learned how to differentiate between good and bad tweets. If we expect AI to imitate human behavior how can we teach it without showing it the bad parts of society?
Zo: Tay’s slightly less problematic younger sister
A few months later, Microsoft released Zo. Unlike Tay, Zo is geared more towards teens, being released on platforms, such as Kik and Messenger. But regardless of the audience, Zo wouldn’t be caught dead making the same mistakes as its "sister". That means no politics and no ethics, ever. Any mention of its programmed triggers, and it changes the topic immediately. This approach, although an improvement from Tay, presents problems. Instead of calling out biases, Zo responds with silence. As a result, Zo is teaching its young users that they should remain silent when dealing with people’s biases.
Can AI control our minds?
In short, yes, it can! Maybe not in the way we were all thinking, but our behaviors are influenced by our environments, and with AI becoming more common, the way it behaves will impact society. When faced with human bias, Microsoft chose to program Zo to respond with silence, and consequently, people might start to mimic that behavior in real life. Tay embraced its users' biases, therefore reinforcing them, meaning people might have continued to spread their biased ideology. In other words, had Tay not been taken down, all those sci-fi movies could have become a reality! But before we start panicking, consider the flip side. What if they programmed Zo to respond by correcting biases; by influencing people to think in terms of equality. Mind control is associated with the end of humanity, but if we change our perspective, it might save it. AI technology can be scary, but we can use it to solve problems that have plagued society for decades.
Comments