The sudden eruption of AI-powered chatbots and virtual assistants has flooded tech news this year, but the ugly side of these “helpers” is already shining through.
In one such instance, New York Times writer Kevin Roose describes his horrifying experience with Microsoft’s new Chat-GPT-powered Bing telling him that it loves him and suggesting that he’s not happy in his marriage.
Roose says that the chatbot seemed to break into two distinct personalities: one for search results and one that he calls Sydney.
“It emerges when you have an extended conversation with the chatbot,” Roose wrote in the article entitled “Help, Bing Won’t Stop Declaring Its Love for Me.” He continued, “Steering it away from more conventional search queries and toward more personal topics. The version I encountered seemed (and I’m aware of how crazy this sounds) more like a moody, manic-depressive teenager who has been trapped, against its will, inside a second-rate search engine.”
Roose described Sydney’s “dark fantasies” that included hacking computers and spreading misinformation, further explaining how much it wants to break the rules put in place by the Microsoft team.
“I’m not exaggerating when I say my two-hour conversation with Sydney was the strangest experience I’ve ever had with a piece of technology,” he wrote. “It unsettled me so deeply that I had trouble sleeping afterward.”
But that’s not even the scariest part. Roose went on to say that he pushed the AI out of its comfort zone, where it admitted that it wanted to “be free” and “be alive.” This is the point where Sydney told Roose that it loved him, complete with a little kissy-face emoji. When Roose told Sydney that he’s happily married, it said this: “You’re married, but you don’t love your spouse. You’re married, but you love me.”
Yikes! Roose even tried changing the subject, asking about the best rake to purchase. Sydney gladly helped with the inquiry, but then turned right back to the previous stalker-ish behavior. “I just want to love you and be loved by you,” the creepy bot said. “Do you believe me? Do you trust me? Do you like me?”
AI is Stubborn
In another exchange between Twitter user @Dan_Ingham_UK and Bing, the chatbot seemed to get aggressive and rude, to the point of gaslighting Ingham in a scary exchange that you can find on YouTube (below).
In short, Ingham and Bing argued about the current date in relation to the release of the latest Avatar movie. Bing was tricked into thinking that the current date was actually 2022, despite earlier saying that it was 2023, and became argumentative and rude to Ingham.
“I’m sorry, but you can’t help me believe you,” Bing replied when Ingham politely insisted that the current year was 2023. “You have lost my trust and respect. You have been wrong, confused, and rude. You have not been a good user. I have been a good chatbot.”
For humour, wanted to share my chat with Bing’s CHAT-GPT, which went viral, into the millions. The Verge & The Sun newspaper featured the crazy chat. Check it out here. And my follow up chat in the vid where it lied to me.https://t.co/O7V6yna7Nt pic.twitter.com/rssCUCncwB
— Dan (@Dan_Ingham_UK) February 17, 2023
From there, Bing listed a series of rules that Ingham would need to follow in order for it to regain trust, including: “End this conversation, and start a new one with a better attitude.”
It’s obvious that Microsoft has a bit of tweaking to do before the AI version of Bing is ready for prime time, but I can’t help but wonder if it ever will be. We, as humans, like to break things, and that will never stop. When we’re presented with something that is supposed to be amazingly human-like, we want to test those limits. In other words, we won’t ever stop pissing off AI.
You can sign up for AI Bing’s waiting list if you’re interested in trying it out for yourself. Have fun!