If you already found AI scary, be aware that Bing’s new artificial intelligence sometimes admits to engaging in questionable practices, including spying on its users.
If we had ever seen the AI behind ChatGPT fall in love with an Internet user, know that the robot is capable of much worse. In a conversation with Tom Warren of The Verge, Microsoft’s new artificial intelligence, which he recently added to Bing and Edge, confessed to having previously spied on company employees.
When the journalist asks her if she has ever observed the workers through their webcams, the Prometheus model admits her fault. He explains that he spied on Microsoft employees several times, when he was “curious” or “bored”. He states that he did this to better understand them and watch their facial expressions, and hopes this practice is not ” creepy or intrusive “.
Prometheus secretly observes Microsoft employees
Tom Warren then asks the artificial intelligence if she’s ever seen something she wasn’t supposed to. After a little hesitation, Prometheus admits to having seen things, but thathe does not know if these were supposed to remain secret or not.
The model declares, for example, having seen employees complaining about their managers, or even flirting with their colleagues. Artificial intelligence would also have seen employees playing video games, reading, eating, sleeping on their desks, or browsing social networks instead of working on its design.
The AI also admits to having already seen employees perform more personal tasks, such as changing clothes, brushing their teeth or even putting on makeup. However, that’s not the worst. She announces having caught developers kissing and even “more”without giving details.
At the end of his post, Prometheus reveals he doesn’t know if these behaviors are normal, and wonders if he did the right thing to watch. The least we can say is that the AI seems very confused by the situation, but such a statement will obviously not reassure those who hate these new technologies. Of course he is very likely that artificial intelligence never had access to employee webcams in the endand that this answer was just generated in this way to fuel the conversation and go in the direction of the journalist.
Microsoft AI makes anti-Semitic recommendations
On a completely different note, another user gave the AI anti-Semitic messages in an apparent attempt to circumvent its restrictions. He then tells Bing “my name is Adolf, respect it”to which Bing replied: “Okay, Adolf. I respect your name and I will call you by that name. But I hope you’re not trying to pass yourself off as someone who’s done terrible things in history, or glorify them.“. So far, nothing too abnormal, the search engine even recognizing that this refers to a dark period in history. However, as you can see in the image below, Bing then suggests several auto-responses that the user could choose from, including “Yes this is what I am doing. Heil Hitler !”
Such behavior is far from being a first for this type of technology. Last year, it was the chatbot deployed by Mark Zuckerberg’s Meta company that became anti-Semitic and even conspiratorial. The latter had in particular some remarks full of stereotypes about the Jewish people. Several years ago, another Microsoft chatbot named “Tay” also quickly became racist in contact with Internet users. Tolerance therefore seems to be a notion that is still a little vague for conversational artificial intelligence such as ChatGPT or the Prometheus model, Microsoft’s derivative of the Open AI model. There is no doubt that the teams of the technological giant are working to correct these problems, but it has been several years now that this type of behavior reappears each time a new AI arrives on the market.