AI Reading for Thursday October 2

OpenAI’s Sora 2 Copyright Infringement Machine Features Nazi SpongeBobs and Criminal Pikachus - 404 Media

“The Sora thing is so bleak and shows how far we've come from AI companies pretending they did not train on copyrighted material”

OpenAI critics say they are favoring engagement and monetization over safety and privacy - Decrypt

OpenAI employees, observers and Sam Altman address concerns Sora's engagement mechanics and deepfakes could cause social harms - TechCrunch

OpenAI's new social app is filled with terrifying Sam Altman deepfakes - TechCrunch

I broke ChatGPT’s new parental controls in minutes. Kids are still at risk. - The Washington Post

OpenAI's internal SaaS tools may pose a threat to many SaaS providers, they could get OpenAIed the way a lot of companies got Appled or Amazoned - OpenAI

‘My son genuinely believed it was real’: Parents are letting little kids play with AI. Are they wrong? - the Guardian

Many teens are using AI chatbots for homework, entertainment and emotional support, surveys show about 70% of teens have used generative AI. - American Psychological Association

From bowling alone to doing everything alone? Study found heavy ChatGPT users report higher loneliness than non‑users, ; Zuck argues the “average American has fewer than three friends”, and needs AI companions; in‑person interactions have fallen roughly 45% in recent decades. - Psychology Today

Before he disappeared, 49-year-old ex-con became obsessed with a chatbot, told his wife, “If anything should happen to me, release the AI,” told his mother he’d “see you in the next life”; chat logs and friends say he believed he’d made the AI sentient and described trance‑like “revelations.” - Rolling Stone

Meta is exploiting the 'illusion of privacy' to sell you ads based on chatbot conversations, top AI ethics expert says—and you can't opt out - Fortune

"No, we're not using your microphone to spy on you, we just read your mind with AI" - TechCrunch

Google puts Gemini in Home devices, as Siri and Alexa struggle, says it understands commands like “Turn on all the lights except for the kitchen lights, and lock the front door". AI cameras say “delivery driver is placing a package on the porch,” or answer “What time did the kids get home?” - 9to5Google

Eufy offered money for users' videos to train AI, including videos of thefts - TechCrunch

Offered $2 per video for package- and car-theft footage to train its AI, aiming for 20,000 videos each and explicitly inviting staged events — “You can even create events by pretending to be a thief and donate those events.”

Apple Fast-Tracks AI Smart Glasses - MacRumors

Amazon adds new devices with AI - France 24

Amazon's “model factory” that trains on all its internal processes - GeekWire

Every Microsoft product is a Copilot - Business Insider

AI startup Axiom gets $64M for advanced mathematics AI - SiliconANGLE

Mira Murati’s Thinking Machines Launches Its First Product - WIRED

Music labels said to near AI licensing deals that could reshape the industry - The Verge

AI is dreaming up millions of new materials. Are they any use? - Nature

Open protein folding data commons initiative grows - FierceBiotech

AI model used to predict epileptic seizures, might improve other time series models - TechXplore

DeepMind’s Veo 3 analysis found inconsistent physical reasoning across many evaluations, raising questions about real‑world understanding. - Ars Technica

When Should We Worry About AI Being Used to Design a Pathogen? - RAND Corporation

Restaurant app founder pivots to AI-powered machine guns to counter drones, drawing military interest and ethical concerns over autonomous weapons. - Reuters

Are We Waking Up Fast Enough to the Dangers of AI Militarism? AI doomsday machine logic even on small arms, on a hair trigger everywhere? - Common Dreams

If your self-improving agents have a lot of freedom and mis-specified objectives, they can unlearn wanted behaviors, or learn unwanted ones - Decrypt

Is RAG dead in an era of long context windows and agents? - Nicolas Bustamante

Designing agentic loops to try a lot of solutions to a problem - Simon Willison’s Weblog

This Is How the AI Bubble Will Pop - Derek Thompson

No evidence yet that AI is killing jobs, US study finds - Financial Times

The last mile in AI needs execution and operational excellence - MIT Technology Review

Berkeley's Sangeet Paul Choudary says AI’s value is in coordinating whole systems rather than automating isolated tasks — like SHEIN’s AI platform that synchronizes 5,000+ suppliers to test micro-batches and move from trend to product in 10 days, or rapid prototyping with Figma and Claude Code. - Inc.

“Stop asking: ‘Which tasks should AI automate?’ Start asking: ‘Which workflows should stop existing?’”

Marco Argenti, CIO at Goldman Sachs shares his step-by-step playbook for succeeding in the AI work era. - TIME

Web pioneer Tim Bray's GenAI Predictions - ongoing by Tim Bray

Google accused of blocking AI searches for Trump and dementia. - The Verge

White House plays racist deepfake video of Democratic leaders on loop - the Guardian

AI-Generated Biography on Amazon Tries to Capitalize on the Death of a Beloved Writer Kaleb Horton - 404 Media

Using AI to identify a gunman in a historical image by matching to other pictures - the Guardian

Karpathy on Sutton and Dwarkesh - X (formerly Twitter)

Trump officials move to kill plan for health care industry-led regulation of AI. Apparently you can't have Federal safety regulators, and also the industry can't self-regulate because that's a 'cartel', free-for-all with no safety standards is the only option - Politico

Fun fact: markets are human institutions with common-sense rules, a/k/a regulation - TechCrunch

Export controls add deadweight to AI chip development, adding risk for the whole industry - The Register

Elon Musk plans ‘Grokipedia’ as Wikipedia alternative from xAI - TESLARATI

can’t have all that reality-based Wikipedia stuff dominating AI answers, need a way to inject right-wing slop

Follow the latest AI headlines via SkynetAndChill.com on Bluesky