All posts tagged: Hallucinations

An encyclopedia formed from AI hallucinations – what could go wrong?

An encyclopedia formed from AI hallucinations – what could go wrong?

Feedback is New Scientist’s popular sideways look at the latest science and technology news. You can submit items you believe may amuse readers to Feedback by emailing feedback@newscientist.com Just a hallucination The online encyclopedias are proliferating. While Wikipedia still dominates, there are plenty of others, like the spectacularly nerdy Memory Alpha, which contains all you could ever want to know about Star Trek. Elon Musk has Grokipedia, a partly AI-generated site that purports to correct Wikipedia’s supposed biases, and in doing so is frequently incorrect. Into this fray enters Halupedia. It is truly unique: it is 100 per cent AI-generated and all of the entries are hallucinations. If you request an article, the site will generate it and then store it indefinitely. Nothing on Halupedia is accurate, except by accident. Hence the site has a page for “The Great Pigeon Census of 1887“, apparently “an ambitious, if ultimately misguided, undertaking by the Royal Society for Avian Enumeration (RSFE) to meticulously count every gold-crested rock dove within the administrative boundaries of the United Kingdom of Great …

Academics in Meltdown Now That They’re Responsible for AI Hallucinations in Their Research Papers

Academics in Meltdown Now That They’re Responsible for AI Hallucinations in Their Research Papers

Sign up to see the future, today Can’t-miss innovations from the bleeding edge of science and tech Even in 2026, there are still plenty of researchers who refuse to use AI to publish their research papers. Others do use the tech for tasks like sourcing journal articles for references, editing copy, or formatting citations — but they face pressure to verify every claim, since AI has a baked-in risk of contaminating their work with hallucinations. A vocal minority of academics, however, argue they should be able to use AI to write original research while remaining immune from any hallucinated claims or data that make their way into the final product. Last week, the open-source research repository arXiv announced that it was banning scholarly authors from the platform for up to a year if “hallucinated references” are found in their work. The rationale behind this should be obvious enough for any self-respecting academic: as arXiv computer science chair Thomas Dietterich wrote in his announcement, “if a submission contains incontrovertible evidence that the authors did not check …

New Wikipedia Clone Made Entirely of AI Hallucinations

New Wikipedia Clone Made Entirely of AI Hallucinations

Sign up to see the future, today Can’t-miss innovations from the bleeding edge of science and tech A new Wikipedia-style site is purportedly made entirely of AI-hallucinations, treating visitors to preposterous insights beamed from a nonexistent reality. Called “Halupedia,” its creators say that the “infinite” encyclopedia invents everything it contains on the fly, with each search term — or link click — becoming a prompt for an AI model on the backend, which relates information “in the deadpan register of a 19th-century scholarly press.” “Every link leads to an entry that does not exist yet — until you click it,” reads the description on GitHub. The site’s homepage is upfront that it’s an exercise in AI fabulation, but once you dive into one of its countless entries, it begins to feel like a real knowledge database, at least if you suspend your disbelief for the many absurdities hurled your way. There’re links, citations, and quotes from academic journals. Some even have footnotes — which, of course, are also made up. One of the top articles is about …

From LLMs to hallucinations, here’s a simple guide to common AI terms

From LLMs to hallucinations, here’s a simple guide to common AI terms

Artificial intelligence is a deep and convoluted world. The scientists who work in this field often rely on jargon and lingo to explain what they’re working on. As a result, we frequently have to use those technical terms in our coverage of the artificial intelligence industry. That’s why we thought it would be helpful to put together a glossary with definitions of some of the most important words and phrases that we use in our articles. We will regularly update this glossary to add new entries as researchers continually uncover novel methods to push the frontier of artificial intelligence while identifying emerging safety risks. Artificial general intelligence, or AGI, is a nebulous term. But it generally refers to AI that’s more capable than the average human at many, if not most, tasks. OpenAI CEO Sam Altman recently described AGI as the “equivalent of a median human that you could hire as a co-worker.” Meanwhile, OpenAI’s charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.” Google DeepMind’s understanding differs slightly from these two definitions; …

Scientists identify brain regions associated with auditory hallucinations in borderline personality disorder

Scientists identify brain regions associated with auditory hallucinations in borderline personality disorder

Recent research published in Psychiatry Research: Neuroimaging suggests that people with borderline personality disorder who hear voices show distinct structural differences in certain brain regions. These physical brain differences tend to appear in areas involved in language processing, sensory integration, and emotional regulation. The findings provide evidence that hallucinations might arise from partly shared biological mechanisms across different mental health conditions. Borderline personality disorder, or BPD, is a severe mental health condition characterized by deep difficulties with self-image, unstable relationships, and challenges in managing emotions. Beyond these well-known symptoms, up to half of individuals with this diagnosis also experience psychotic symptoms. Specifically, many of these patients experience auditory verbal hallucinations, which means they hear voices that other people do not hear. Historically, the scientific focus on auditory hallucinations has centered on schizophrenia. Because of this, the physical brain characteristics linked to hearing voices in borderline personality disorder have remained poorly understood. Scientists wanted to see if specific brain structures differ in people with borderline personality disorder who hear voices compared to those who do not. …

ChatGPT Gets GPT-5.3 Instant Update With Less ‘Cringe,’ Fewer Hallucinations

ChatGPT Gets GPT-5.3 Instant Update With Less ‘Cringe,’ Fewer Hallucinations

OpenAI today updated its most popular ChatGPT model, debuting GPT-5.3 Instant. GPT-5.3 Instant is supposed to provide more accurate answers and better contextualized results when searching the web. The update also cuts down on unnecessary dead ends, caveats, and overly declarative phrasing, plus it has fewer hallucinations. According to OpenAI, it tweaked the Instant model to address complaints about tone, relevance, and conversational flow, which are issues that don’t show up in benchmarks. GPT-5.2 Instant had a “cringe” tone that could be overbearing or make unsubstantiated assumptions about user intent or emotions. The new model will have a more natural conversational style and will cut back on dramatic phrases like “Stop. Take a breath.” Users found that GPT-5.2 Instant would refuse questions it should have been able to answer, or respond in ways that felt overly cautious around sensitive topics. GPT-5.3 Instant cuts down on refusals and tones down overly defensive or moralizing preambles when answering a question. The model will no longer “over-caveat” after assuming bad intent from the user. GPT-5.3 Instant also provides …

GPT-5.3 Instant cuts hallucinations by 26.8% as OpenAI shifts focus from speed to accuracy

GPT-5.3 Instant cuts hallucinations by 26.8% as OpenAI shifts focus from speed to accuracy

OpenAI’s GPT-5.3 Instant — the company’s most widely used model — reduces hallucinations by up to 26.8% compared to its predecessor, prioritizing accuracy and conversational reliability over raw performance gains, OpenAI says. GPT-5.3 Instant, which is essentially the default and is the most used model for ChatGPT users, also improves on tone, relevance and conversation with fewer refusals. It is available on both ChatGPT and on the API.  Right now, only the Instant model will be upgraded to 5.3, but the company said it is working on updating the other models under ChatGPT, Thinking, and Pro to 5.3 “soon.”  GPT-5.3 Instant cuts hallucinations by up to 26.8% OpenAI ran two internal evaluations: one across higher-stakes domains including medicine, finance, and law; the other drawing on user feedback. Based on higher-stakes evaluations conducted by the company, GPT-5.3 Instant reduces hallucinations by 26.8% when using the web. It improves reliability by 19.7% when relying on its internal knowledge. User feedback showed a 22.5% decrease in hallucinations when answering queries using web search.  The company said GPT-5.3 Instant …

Spot AI Hallucinations Like a Reference Librarian and More Library News

Spot AI Hallucinations Like a Reference Librarian and More Library News

This content contains affiliate links. When you buy through these links, we may earn an affiliate commission. Prep those holds lists and get your selector carts ready because the new year means all kinds of new book and book adaptation announcements! I’ve got lots of links to help library workers keep their collection development game strong, plus helpful articles, like what to do if your library falls victim to a ransomware attack, and how to spot AI hallucinations “like a reference librarian.” Let’s dive in. New & Upcoming Titles Patricia Cornwell has a new memoir coming out in 2026. Jesmyn Ward has a new nonfiction title coming out this spring. Jason & Travis Kelce have a book coming out this summer. Here’s the cover reveal for Amy Bloom’s upcoming mystery novel, Blunt Instrument, which comes out on June 2nd. A new Heated Rivalry book is coming. 20 of the best new reads for your book club. 2026 picks from BBC, Crime Reads, LitHub, New York Times (fiction, nonfiction), NPR, StarTribune, Town & Country (romance), Vulture, …