AI Reading for Wednesday August 20

Sam Altman just teased GPT‑6 with more personalization. - Tom's Guide

If GPT-5 is PhD level (doesn't feel that way) what's next, Einstein level?

The Day ChatGPT Went Cold - The New York Times

Gmail and Google Calendar connectors arrive for Plus and Pro users, enabling inbox and schedule queries. - TechRadar

Gmail already has native AI search, like, find that story I got from Skynet And Chill enumerating business models for AI startups

OpenAI CEO Sam Altman says that export controls won’t hold back China's AI ambitions — “My instinct is that doesn’t work” - Tom's Hardware

After recent tests, China appears likely to beat the United States back to the Moon - Ars Technica

Chinese Livestreaming 'Virtual Human' Salespeople Are Outselling Their Human Counterparts - 404 Media

China pushes widespread AI adoption in business, reframing competition as human-with-AI versus human. – - Marketplace

I was in China in 2010 and felt like a Parisian or Londoner would have felt in 1900 visiting NYC or Chicago, they were 10 years ahead on cashless society, getting everything by electric delivery scooters, insane infrastructure. Their Trump (Xi) is also a smarter and more evolved model.

Cursor agent mode and Claude Code cause existential crisis for a software dev - Hacker Noon

GitHub workflows to automate dev with Claude Code integration - GitHub

Xcode teardown shows Claude coding assistant support - Ars Technica

The weird thing about this feature where Claude will now end the most unsafe conversations is, it came from a place of research into 'model welfare', which I'm not sure they have shown is actually a thing. If you ran the model with paper and pencil, would the model have feelings? - Anthropic

xAI made conversations with Grok searchable on Google without user consent when they were shared, including a plan to assassinate Elon Musk and instructions for making fentanyl and bombs. - Forbes

Lawsuits test Tesla's self-driving claims - Los Angeles Times

ARM poached Amazon’s AI chip director to lead its in-house chip push, expanding beyond IP licensing - Yahoo Finance

SoftBank held talks with Intel on buying contract chipmaking business - Financial Times

Meta rolls out AI-powered translations, in your own voice, to creators globally, starting with English and Spanish - TechCrunch

Meta is restructuring its AI division into four groups, exploring third-party models, and may cut thousands of jobs - The New York Times

How Meta Became Uniquely Toxic for Top AI Talent. Mark Zuckerberg’s nine-figure job offers were just one of multiple bad signs. - Intelligencer

Internal Meta documents reportedly allowed generation of racist and false medical content, fueling policy and safety concerns - Futurism

Is Google behind a mysterious new nano-banana AI image generator? These bananas might confirm it. - Business Insider

Google inks a deal with Kairos and TVA for SMR nuclear power - Interesting Engineering

Oracle powers its way into AI business but at the price of negative cash flow, being on the hook for future billions - Bloomberg

How AI-enhanced hackers are stealing billions - The Economist

Scammers are injecting fake customer-service numbers into Google AI Overviews and ChatGPT responses, tricking users into fraud - ZDNET

Research warns AI-boosted velocity is outpacing testing, raising outage risks and tech debt. - TechRadar

US AI deregulation could lead to chaos, cede regulatory framework to Europe. - The Register

Americans fear AI permanently displacing workers, Reuters/Ipsos poll finds, as well as political manipulation - Yahoo Finance

A randomized trial found AI-led interviews improved call center hiring outcomes and 30-day retention versus humans. - Business Insider

Teachers increasingly use AI for planning and productivity, while tech firms fund large‑scale training and unions call for guardrails - CBS News

Kellanova tests AI to measure Pop-Tart creative effectiveness and reshape agency fee models, to incentivize more attention-getting and effective creative. - Digiday

Microsoft Thinks AI Can Replace Wealth Managers. Advisors Say: Yeah, Right. - The Daily Upside

KPMG wrote 100-page prompt to build agentic TaxBot - The Register

Delivers in hours what a tax advisors might take days to research.

Human OnlyFans chat assistants who impersonate then on-air talent face replacement by LLM-trained bots as agencies automate chat at scale. - Rest of World

Runway AI paired up with Imax to screen 10 AI-generated winning selections. The festival has been derided by some cinephiles, while others insist it’s part of a natural technological evolution. - WIRED

Bill Gates-backed AI competition offers $1M prize for best use of agentic AI to accelerate Alzheimer’s research - GeekWire

A machine learning model on lifestyle and DNA claims very high accuracy for cancer risk prediction, cautious deployment proposed. - Nature

Epic is launching an AI scribe tool across its platform, expected to pressure startup pricing in the $1B ambient AI market - Fierce Healthcare

AI tools found useful for summarizing, not so much for journalist’s research - www.cjr.orgTrump, Putin and European leaders become AI slop, meme, and misinformation fodder - The Moscow Times

Follow the latest AI headlines via SkynetAndChill.com on Bluesky