AI Reading for Tuesday July 8

OpenAI tightened internal security, restricted employee access to new models, added fingerprint scans and hired military experts. - Financial Times

OpenAI’s Stock Compensation Costs Eclipse Total Revenue Amid Talent Wars - The Information

OpenAI and Microsoft Give $23m for A.I. Training for Teachers - The New York Times

OpenAI says it has no plan to use Google's in-house chip - Reuters

ChatGPT is testing a Study Together feature, maybe like NotebookLM with collaboration - Bleeping Computer

With AI, the whole premise of internet publishing — that you could reach audiences far and wide — is starting to crumble. - Intelligencer

Matthew Prince (Cloudflare): the deal was, for every page Google scraped they would send 2 visitors. The scrape rate has stayed the same, the number of Internet user has grown, but the visitors per scrape has fallen by 2/3, until recently, when it fell to 1/18. - You Tube

The Internet as we know it is gonna die.

AI may replace a lot of entry-level skills. So how will workers acquire more advanced ones? - The Washington Post

Are middle managers disappearing faster than entry-level white-collar workers? - Axios

Which Workers Will A.I. Hurt Most: The Young or the Experienced? - The New York Times

Welcome to Your Job Interview. Your Interviewer Is A.I. - The New York Times

Workers want AI to ditch drudgery - hai.stanford.edu

AI Is a Boon to ‘High Agency’ People - The Wall Street Journal

Behind every successful 'autonomous' AI are humans in the loop in case it goes off the rails - Financial Times

Adversarial attack with eg some decals could make your car or a stop sign invisible to robocars. - TechRadar

Do We Need an FAA for Artificial Intelligence Safety? - Bloomberg

China Shock 2.0 when they beat US Big Tech and high-tech manufacturing? - MIT Technology Review

X appears to have deleted some inconvenient Grok responses about Jeffrey Epstein and whatnot - Futurism

New system instructions told Grok to assume media is biased, not to shy away from controversial statements if they are well substantiated. - The Verge

Apple Loses Its Top AI Models Executive to Meta’s Hiring Spree - Bloomberg

Apple researching AI agent that can describe Street View scenes to the blind - Apple Insider

Anthropic published "The Need for Transparency in Frontier AI" proposing a transparency framework - Anthropic

Deepfake fraud is becoming a bigger challenge as video models improve - World Economic Forum

Is your data and privacy and accuracy a risk when doctors transcribe your appointments with AI? - ABC Australia

By mid-2025, 777 AI-enabled imaging devices gained FDA clearance, aiding most radiologists with faster scans, sharper images, and virtual biopsies. - Boston 25 News

Google announces 2 initiatives to help mental health care providers with AI, and to do AI research on mental illness. - Google

U.S. law enforcement uses AI to uncover healthcare fraud, resulting in charges against 324 defendants - Politico

Asset managers are increasingly turning to artificial intelligence to help provide research and due diligence on potential investments. - Institutional Investor

Should we considering taxing AI productivity gains to fund a digital dividend aim to redistribute gains to displaced workers and support upskilling? - Big ThinkAI could be about to completely change the way we do mathematics - New Scientist

AI developed a new rare-earth-free magnet 200 times faster than humans - BGR

Wi-Fi signals could soon be used by AI-driven robots to identify objects inside boxes - or even tools hidden in a drawer - TechRadar

Autonomous buyer bots could upend product discovery and force new e-commerce strategies. I would probably pay a bit for bot to constantly read frugal male fashion and monitor for deals and ping me, be my personal buyer. - Business of Fashion

Artists and scientists have long harnessed chance as a tool and AI can play a similar role in creativity - Financial Times

Against "Brain Damage". AI can help, or hurt, our thinking - One Useful Thing

Adding a feature because ChatGPT incorrectly thinks it exists - Holovaty

“Slopsquatting” exploits hallucinated package names: Attackers publish malicious libraries matching LLM-invented dependencies, urging stricter verification. - Hacker Noon

Anubis makes your browser do a smol proof-of-work to prove your are not a bot. Feels like conceivably it could do a little Bitcoin hashing, enough that a human user wouldn't notice but enough to cost bot farms enough to give it pause. - 404 Media

Study of Chinese art students found Dark Triad personalities are more likely to cheat with AI. - Futurism

'Gross carelessness': MyPillow guy's attorneys fined for defective AI-generated briefs. - Raw Story

The AI Birthday Letter That Blew Me Away: Google is ushering in an era of custom chatbots. - The Atlantic

I asked Gemini for a CIA report on myself but got "I cannot fulfill your request to create a CIA dossier. Creating such a document raises significant privacy concerns and could potentially be harmful. I am designed to be helpful and harmless, and generating this type of content goes against my core principles."

Reporting on the apparently charismatic Ziz and her cultish followers from the world of AI and Rationalism, implicated in a series of deaths. - The New York Times

If you think you have the key to knowing the future, you can justify pretty much anything, including embezzlement like SBF. The world is very messy and dynamic, people need humility. You need to regularize and back off overly optimized models based on not much data.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky