Search Engine Journal: AI Researchers Warn: Hallucinations Persist In Leading AI Models. “Despite billions in research investment, AI factuality remains largely unsolved. According to the report, even the most advanced models from OpenAI and Anthropic ‘correctly answered less than half of the questions’ on new benchmarks like SimpleQA, a collection of straightforward questions.”

https://rbfirehose.com/2025/04/01/ai-researchers-warn-hallucinations-persist-in-leading-ai-models-search-engine-journal/

Poynter: Opinion | An Italian newspaper launched a generative AI experiment. It’s not going well. “I am usually a fan of bold experiments like this, and bristle at the hand-wringing that often comes in response to emerging technology. But, so far, it’s a case study of how bad AI is at writing the news. Along with the misspellings in images, an article about Donald Trump’s falsehoods has […]

https://rbfirehose.com/2025/03/23/opinion-an-italian-newspaper-launched-a-generative-ai-experiment-its-not-going-well-poynter/

TechCrunch: ChatGPT hit with privacy complaint over defamatory hallucinations. “OpenAI is facing another privacy complaint in Europe over its viral AI chatbot’s tendency to hallucinate false information — and this one might prove tricky for regulators to ignore. Privacy rights advocacy group Noyb is supporting an individual in Norway who was horrified to find ChatGPT returning made-up […]

https://rbfirehose.com/2025/03/21/techcrunch-chatgpt-hit-with-privacy-complaint-over-defamatory-hallucinations/

TechCrunch: Manus probably isn’t China’s second ‘DeepSeek moment’. “Alexander Doria, the co-founder of AI startup Pleias, said in a post on X that he encountered error messages and endless loops while testing Manus. Other X users pointed out that Manus makes mistakes on factual questions and doesn’t consistently cite its work — and often misses information that’s easily found […]

https://rbfirehose.com/2025/03/11/techcrunch-manus-probably-isnt-chinas-second-deepseek-moment/

NiemanLab: AI search engines fail to produce accurate citations in over 60% of tests, according to new Tow Center study. “Last summer, I reported that ChatGPT frequently hallucinated fake URLs to news sites, even to articles from OpenAI’s own publishing partners. Research has continued to show that these citation issues are not limited to ChatGPT, but are in fact chronic across the AI […]

https://rbfirehose.com/2025/03/11/niemanlab-ai-search-engines-fail-to-produce-accurate-citations-in-over-60-of-tests-according-to-new-tow-center-study/

BBC: Grandmother gets X-rated message after Apple AI fail. “Louise Littlejohn, 66, received a voicemail message on Wednesday from a Lookers Land Rover garage in Motherwell inviting her to an event. An artificial intelligence (AI) powered service offered by Apple turned it into a text message which – to her surprise – asked if she been ‘able to have sex’ before calling her a ‘piece of ****’.”

https://rbfirehose.com/2025/03/09/bbc-grandmother-gets-x-rated-message-after-apple-ai-fail/

Key Points:
AI errors running amok like a robot revolution, featuring Google search gaffes and Comcast calendar calamities.
Flubstitution: when AIs confuse template placeholders with Mad Libs.
Case study in how open-source session recording gets downright mysterious.
Dell's UI patterns? More like antipatterns.

https://thedailywtf.com/articles/artificial-average-intelligence

Hey, here's Copilot being all inaccurate on searches I did about how long journals should retain copies of a peer review. #AIErrors

»#Anthropic's new #Citations feature aims to reduce #AIerrors: allows its #AImodels to provide references to “the exact sentences and passages” from docs they use to generate responses.« https://techcrunch.com/2025/01/23/anthropics-new-citations-feature-aims-to-reduce-ai-errors/?eicker.news #tech #media

Washington State University: ChatGPT errors show it cannot replace finance professionals, yet. “While large language models like ChatGPT can do well when choosing multiple-choice answers on financial licensing exams, they falter when dealing with more nuanced tasks. A Washington State University-led study analyzed more than 10,000 responses to financial exam questions by the artificial […]

https://rbfirehose.com/2024/12/19/washington-state-university-chatgpt-errors-show-it-cannot-replace-finance-professionals-yet/

ChatGPT goes temporarily “insane” with unexpected outputs, spooking users - Enlarge (credit: Benj Edwards / Getty Images)

On Tuesday, Chat... - https://arstechnica.com/?p=2004783 #largelanguagemodels #machinelearning #aitransparency #openweightsai #textsynthesis #gpt-4-turbo #aierrors #aisafety #bingchat #chatgpt #chatgtp #gpt-3.5 #biz⁢ #openai #gpt-4 #api #ai