AI Reading for Saturday August 23

AI slop is bad - Snopes

I sense some frustration with fact-checking this stuff and seeing results like this when doing research

AI‑generated images of injured Israeli soldiers spread widely, fueling propaganda and donation scams. - France 24

Fans loved her new album. The thing was, she hadn't released one. Artists report AI‑generated albums falsely released under their names on streaming platforms, exposing gaps in verification and controls across music distribution. - BBC

AI‑driven ‘ghost student’ fraud rings exploited college systems to siphon federal aid, prompting new identity checks and AI defenses, harder look at access‑equity trade‑offs. - Fortune

AI Helps UK Woman Rediscover Lost Voice After 25 Years - International Business Times

An Arizona sheriff’s department used ChatGPT to generate a photo from a sketch, helping identify a deceased man during an ongoing investigation. - FOX 10 Phoenix

Sen. Klobuchar sets record straight: She never said “If Republicans are gonna have beautiful girls with perfect t-tties in their ads, we want ads for Democrats too, you know?” - New York Post

Nvidia Readies New China Chip as Washington Debates A.I. Exports - The New York Times

Google says it dropped the energy cost of AI queries by 33x in one year - Ars Technica

Big AI training jobs have power demand swings large enough to impact the grid when they pivot from GPU use to redistributing data between GPUs - The Register

Amazon is betting on agents to win the AI race - The Verge

The age of experience supplants the age of innocence - David SIlver

Elon Musk says he wants to 'simulate' software companies like Microsoft 'purely' with AI. He's calling it 'Macrohard.' - Business Insider

Apple preps enterprise management for ChatGPT and potentially other AI providers in iOS 26, with admin controls, APIs, and device‑management improvements. - 9to5Mac

Apple in Talks With Google to Power Next-Gen Siri With Gemini AI - MacRumors

Meta is going to stuff Midjourney AI images into your feed - The Verge

Meta to unveil Hypernova smart glasses with a display, wristband at Connect next month - CNBC

ChatGPT 5 is finally saying 'I don’t know', hallucinations said to be significantly reduced - TechRadar

A study comparing 1,500 AI respondents to real people found large errors, especially for subgroups, undermining the reliability of synthetic polling samples. - Futurism

To dry-run social science research and fix questions, OK but expecting valid results is nuts, it's hard to get good results with real people.

Can AI ‘sorcery’ solve the ‘productivity paradox’ that has gripped the economy for 25 years? A Shakespearean sea change is underfoot - Yahoo Finance

Most social media managers use AI for ideation and drafting but cite quality, bias, and reputational risks requiring human oversight. - CNET

When I let ChatGPT manage my social life for a week, AI texted my friends and picked my outfits. It was strange but helpful. - Business Insider

“Vibe coder” advocates supervising many small LLM agents for refactors and tests, with realistic productivity limits. - Stochastic Lifestyle

DeepSeek v3.1 Is Not Having a Moment - Zvi

benchmarks aren’t everything

YouTube’s is quietly using AI to “improve clarity” - The Atlantic

AI robot dogs deliver fast food in Zurich, as Just Eat pilots new technology - YouTube

The White House ordered federally contracted AI models to certify no “ideological bias,” drawing civil-liberties warnings about censorship and chilled speech. - Techdirt

what does that even mean. part of it apparently is not filtering fake quotes about Sydney Sweeney’s t*tties.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky