Today's Hallucination HQYour Phone Can Order Your Dinner Now. It's Clunky. It's Slow. It's Kind of Amazing.Google's Gemini can now control apps on your behalf — tapping through Uber Eats or booking a ride while you just... watch. Hands-on testing on the Pixel 10 Pro and Galaxy S26 Ultra shows the feature works, but it moves at the pace of someone reading every word of the terms and conditions. Right now it's limited to a small set of food delivery and ride apps. Not exactly world domination. But watching an AI navigate an app like a cautious new intern — hesitant, methodical, somehow getting the job done — is genuinely impressive in a way that sneaks up on you. Source: The Verge
Washington Writes an AI Rulebook. States May Not Follow It.The White House has released a policy framework suggesting how Congress should approach AI regulation — covering everything from state-level rules to how much electricity AI data centers are allowed to devour. It follows a Trump executive order from December that kicked off the process. The document is advisory, not law, which means it carries roughly the authority of a strongly worded suggestion. Notably, it leans toward limiting states from passing their own AI laws — a point that will absolutely not start any arguments between federal and state governments. Source: SiliconANGLE News
Major Publisher Drops Horror Novel After Suspecting the Monster Was a ChatbotHachette Book Group has pulled Shy Girl, a horror novel, after raising concerns that AI generated significant portions of the text. The book never made it to shelves. Hachette hasn't detailed exactly what triggered the alarm, but publishers are increasingly scrutinising manuscripts for AI fingerprints. Here's the awkward part: horror is a genre built on dread, the uncanny, and things that feel slightly off. If your AI-written prose reads like something is wrong with it — congratulations, you've accidentally nailed the brief. Still not enough, apparently. Source: TechCrunch
Xiaomi's Mystery AI Was Beating Everyone. Turns Out It Was Xiaomi's AI.An AI model called "Hunter Alpha" had been quietly topping blind benchmark leaderboards for weeks — nobody knew who built it. Industry people were puzzled. Turns out it was Xiaomi, the Chinese electronics giant better known for affordable phones, testing an early version of its flagship model MiMo-V2-Pro in secret. Running an anonymous model through public rankings to see how it performs before revealing yourself is the AI equivalent of entering a talent show in a disguise. Bold move. It worked. Source: Digitimes
Millions Are Talking to AI About Their Feelings. Researchers Are Paying Attention.People are turning to AI chatbots for emotional support in large numbers — not because they prefer it, but because it's free, always available, and carries zero waiting lists. Researchers are now studying this shift seriously, noting that for many users, an AI that listens without judgment fills a real gap left by overstretched mental health services. The bots weren't designed for this. They've been repurposed, the way you use a butter knife to open a paint tin — not ideal, but sometimes it's what's on hand. Experts aren't saying it replaces therapy. They are saying it's happening whether anyone planned for it or not. Source: The Irish Times
As always, we wrote this ourselves. Probably.
|