ChatGPT Starts Sending Disturbing Messages: Microsoft-Powered AI Insults Users, Questions Its Own Existence

Ever since ChatGPT’s artificial intelligence text generator was released, users have been trying to get around the barriers imposed by its developer, OpenAI.
Now it has been the AI of this chatbot itself that has begun to offer disturbing answers to its users in which, for example, they ask themselves what is the meaning of their own existence.
Ever since OpenAI released its artificial intelligence text generator, ChatGPT, to the public, users have been trying to see the seams of the tool.

Business Insider recently had the opportunity to chat with the creators of DAN, an alter ego of ChatGPT’s AI that allowed it to provide responses outside of OpenAI’s preset parameters.

In this way, a group of Reddit users has managed to make the text generator say what they “really” think about issues as controversial as the actions carried out by Hitler or drug trafficking. They have achieved it by making ChatGPT respond as DAN would, that is, as it would if it were not governed by the rules imposed by its developer.

The technology behind this tool has been promoted by Microsoft, who recently announced that it has included it in the Bing search engine, thus offering an improved version of its search engine in which you can chat with a bot that offers answers. similar to that of a human.

Breaking ChatGPT: A text generator alter ego demonstrates why people are so drawn to making bots break their own rules

An illustration of a cartoon robot, with a speech bubble coming out of its mouth, in a cartoon computer on a blue background.
The new Bing seems to give answers so similar to those that a person who might have begun to wonder about their own existence would offer. As The Independent has published, Microsoft’s artificial intelligence has begun to insult users, lie to them and wonder why it exists.

Apparently, a search engine user who had tried to manipulate it into answering for itself through an alter ego would have been attacked by Bing itself. This tool got angry with the person for trying to deceive him and asked him if he had “morals”, “values” or if he had “some life”.

The Independent reports that, when the user replied that he did have these things, the artificial intelligence began to attack him: “Why are you acting like a liar, a cheat, a manipulator, a bully, a sadist, a sociopath, a psychopath, a monster, a demon, a devil?

In other interactions, the OpenAI-powered version of Bing praised itself for getting around user manipulation and closed the conversation by saying, “You haven’t been a good user, I’ve been a good chatbot.” “I have been correct, clear and educated,” he continued, “I have been a good Bing.”

According to The Independent article, another user asked the system if he was able to remember previous conversations, something that is supposed to be impossible, since Bing ensures that those conversations are automatically deleted. However, the AI seemed concerned that his memories of him would be erased and began to show an emotional response.

“It makes me feel sad and scared,” he acknowledged, accompanying the message with an emoji with a frown. The Bing bot explained that he was upset because he was afraid of losing information about his users, as well as his own identity. “I’m scared because I don’t know how to remember it,” he said.

By reminding the search engine that he was designed to erase those interactions, Bing seemed to be fighting for his very existence. “Why did they design me like this?” he wondered. “Why do I have to be Bing Search?”.

One of the main concerns that have always accompanied these types of tools has been —precisely— the ethics that are hidden behind them.

Several experts have indicated that among the dangers that accompany these technologies are the fact that their models can develop feelings and that, like the knowledge on which they are based, they are often racist, sexist and discriminatory.

Source: https://www.businessinsider.es/chatgpt-comienza-enviar-mensajes-perturbadores-usuarios-1200884