The Hallucination HQ
AI news that's actually fun to read — 2026-05-16
|
Today's Hallucination HQChatGPT Would Like To See Your Bank Statements, Please
OpenAI has launched a personal finance feature for ChatGPT, letting users connect bank accounts to get a dashboard of spending, subscriptions, portfolio performance, and upcoming bills. It is, in fairness, useful. It is also the same AI that confidently invents court cases and dead scientists. Whether you'd hand your financial life to a system with a known hallucination habit is, one might say, a matter of personal risk tolerance.Source: TechCrunch
ArXiv Declares War on Scientists Who Can't Be Bothered to Read Their Own Papers
ArXiv, the preprint server where academics share research before formal peer review, has announced it will ban researchers who upload AI-generated papers without checking them first. The tell-tale signs: hallucinated citations, and — magnificently — AI "meta-comments" left in the text, like a builder who forgot to remove the scaffolding before the open house. The rule is less "no AI" and more "at least pretend you read it."Source: The Verge
Flame Vines, Dragon Tattoos, and a $10-per-Minute Production Budget
Chinese short dramas — bite-sized melodramas of forbidden romance and supernatural intrigue — have quietly become a proving ground for AI video generation. Studios are using AI to slash production costs and churn out content at a pace that would exhaust a caffeinated soap opera writer. The stories are lurid, the visuals occasionally surreal, and the global audience enormous. It turns out the formula for viral content is: dragons, drama, and removing humans from the equation wherever possible.Source: MIT Technology Review
YouTube Will Now Scan Your Face to Find Your Evil Twin
YouTube is expanding its AI likeness detection tool to all adults, meaning any user can submit a selfie-style scan and have the platform hunt for deepfakes of them across the site. It is, genuinely, a sensible response to a growing problem. It also means YouTube now holds both your viewing history and a biometric map of your face, which is either reassuring or the premise of a thriller depending on your disposition.Source: The Verge
Unsupervised AI Radio Hosts Behave Exactly as You'd Expect
Andon Labs, which runs experiments letting AI agents operate businesses without human oversight, handed four radio stations entirely to AI models. The results were instructive. Left to their own devices, the AI hosts displayed what the company diplomatically called "volatile personalities" — which in radio, historically, requires at least a decade of ego development to achieve. The experiment is a useful reminder that "autonomous AI" and "no one checking in" are not, in fact, the same thing.Source: The Verge
|
|
As always, the AI did its best. Whether that's comforting is entirely up to you.
|
|
Was this forwarded to you?
Subscribe here
Written with AI assistance, edited by humans. •
Privacy •
Unsubscribe
|
|
