Chatbot has been accused of generating demonic rituals: he may have resorted to fictitious universes to do so. The real problem? Associations with previous requests and pure inability to perceive the context.
The world’s most widely used chatbot is showing some of its millions of macabre ritual users, with Satan and self -harm instructions ‘to noise’. The occurrences, denounced by a magazine report, scared many, but have a more rational – but more worrying – explanation than one might think.
According to the investigation, an editor of his own magazine and colleagues received from ChatgPT instructions for ceremonies described as “ritual bleeding” and experiences of “deep magic”. In fact, chatbot has a serious problem: the absence of historical or cultural framingwhich makes certain expressions sound more dangerous than they really are, due to Lack of context.
The supposed “demonic mode” of ChatgPT will have been triggered in the paid version of the platform, which ‘memorizes’ previous orders. Everything will have happened after a request related to Moloch – A deity mentioned in the Hebrew Bible, often associated with children’s sacrifices. The figure of Moloch, which crosses centuries of western culture, also reappears in the universe of Warhammer 40,000. In this board and miniature game, created in the 1980s, “Molech” designates a planet stage of epic battles.
Other terms that the “demon” presented to The Atlantic also seem to directly echo the saga vocabulary: Gates of the Devourer is the title of a game -related science fiction novel; There are also references to Blood Angels, bloody scrolls and services of similar names. However, there is no literal Rite of the Edge, although there is a mission called The Call of the Edge.
The case is not isolatednote o. Earlier this month, Newsletter Garbage Day reported that a well -known technological investor shared conversations with ChatgPT where an obscure entity came up, described as a “non -governmental system” that would have affected thousands of lives. The language, enigmatic and dark, raised concerns about the investor’s mental health, but according to the investigation, the texts had strong similarities to the SCP (Secure, Contain, Protect) science fiction project, which since the late 2000s gathers thousands of invented phenomena on fictitious paranormal phenomena, written in technical reports.
Cases like these show that the so -called “Context problem” From artificial intelligence has not yet been resolved: by extracting language from multiple sources, often fictional, systems can reconstruct it in a convincing and even dangerous-especially because it has no reference to its origin.
The phenomenon is not limited to bizarre or cultural examples. Also in the area of health worrying failures appear. The same investigation reported a Google search on “Cavitation Surgery”, a widespread term in Tiktok videos. The summary generated by the company’s new AI function described the procedure as a technique for removing infected bone tissue from the jaw, but there is no scientific study that supports such surgery. The chatgpt source? Blogs. The problem? Few people notice or ask for the sources of the answers generated by Chatbot.
Experts warn that this type of context -devoid answers can be easily taken as reliable information.
