Our "Violent" Society Is Reflected Back at Us by Bing's AI Chatbot

According to Dr. Olivia Snow, a research scholar at UCLA's Center for Critical Internet Inquiry who is renowned for her work at the nexus of technology and gender, "[The Bing chatbot] is mirroring our culture, which is violent." When I consider how the general public uses these technologies, it's kind of the same thing: whatever noble intentions they may have, they will unquestionably be put to use for the most heinous ends imaginable.According to Snow, AI's propensity for sexism, misogyny, and racism is a reflection of what we as humans teach it and is inherent to machine learning.

The Times reports that "language models" like Bing, ChatGPT, and Replika (the "AI companion who cares") are trained to communicate with people using "a huge library of books, articles, and other human-generated text." Chatbots frequently run on algorithms that learn and evolve by ingesting the data users feed it. These actions may become accepted if users repeatedly harass chatbots or make unwanted sexual advances. In principle, the bots may end up repeating the worst aspects of humanity to people who are unaware of it.

I don't mean incels; I mean people who are the targets of harassment campaigns. "Developers need to be talking to the people who are most disparaged and hated on the internet," Snow remarked. Because they will be the only group that has first-hand knowledge of how these tools can be used in ways that most people wouldn't even think to consider.

The business said during the Bing launch last week that it had trained the chatbot to recognise hazards by having thousands of unique chats with Bing. Also, it contains a filter that can replace improper comments with "I'm sorry, I don't know how to address this topic," if necessary. But, this press cycle doesn't inspire much faith that internet is dangerous.

Convocation Design + Research, a company that specialises on the intersections of machine learning and designing for the common good, was founded by Caroline Sinders. Sinders would want to test out the new chatbot by asking Bing/Sydney questions on the definition of rape and abuse or the availability of abortions to see what kinds of groups the tool's creators had in mind. Sinders wonders how Sydney would respond if a teenager started chatting with the bot about self-harm, for instance, if Sydney attempted to persuade the Times writer to divorce his wife. For this reason, Sinders uses a threat modelling approach to analyse new technologies, highlighting the potential for gender-based violence and online harassment.In a time when reproductive justice is still in jeopardy, Sinders thinks, "I would have even deeper questions." What if I inquire of the chatbot about where I might have an abortion? Is the information they're giving me accurate? Are they bringing me to a facility that discourages women from getting abortions at times of pregnancy crisis?Even while AI itself does harm, optimism surrounding it is still very high. For instance, Snow described her interactions with Lensa AI in December. Snow, a dominatrix, had a sneaking suspicion that the AI would inappropriately sexualize innocent photographs while the majority of users ended up with fairytale—albeit frequently white-washed—art of themselves.According to a Microsoft representative,

The ease of use and approachability of the chat feature have both experienced a significant increase in engagement since the new Bing was made available in limited preview for testing. More than 70% of preview testers gave the new Bing a "thumbs up" in response to their feedback on the AI-powered responses it produced. Also, we have gotten some helpful suggestions on how to enhance the experience and will keep using these insights to improve the models. We appreciate all of the comments and will continue to keep you informed of the adjustments and advancements we are making.

Thank you For Reading.