AI Reading for Monday September 15

Behind the crisis in OpenAI and Microsoft's relationship. - aimagazine.com

AWS, Meta, Apple etc. don’t have it together but Google does and will eat their lunch if they can’t sort themselves out

Sam Altman discussed content moderation, privacy aims, possible military links, and near-term job losses, promising improved crisis responses as OpenAI refines ChatGPT safety, also Tucker Carlson asked if he had had anyone murdered lately. - CNBC

OpenAI is assembling a dream team to develop algorithms to control robots and appears to be hiring roboticists who specialize in humanoid robots, signaling a ramp-up in its robotics efforts. - WIRED

OpenAI published the first detailed public study of ChatGPT users and their interactions, "providing an unprecedented look at how people use the artificial intelligence tool and what they talk to it about." - OpenAI

Women are the majority now, and personal use outstrips professional use.

Google’s Gemini app reached No. 1 on Apple’s Top Free Apps chart in the US (ChatGPT No. 2, Threads No. 3); other Google apps in the US rankings include Google Search No. 6, Maps No. 8, Chrome No. 13 and Gmail No. 21; Gemini is No. 2 in Canada and the UK. - 9to5Google

Google DeepMind CEO Demis Hassabis warned that AI must not "repeat the same traps that made social media toxic," urging a more cautious rollout than Silicon Valley's old "move fast and break things" ethos and saying AI should "serve people, not to manipulate them." - Business Insider

Anthropic Economic Index report: Uneven geographic and enterprise AI adoption - Anthropic

AI usage concentrated in high-income countries, rising automation by consumers, and enterprises favoring automation-heavy deployments.

Biggest raises so far in 2025 - Quartz

• OpenAI ($40 billion, more than any IPO ever I think)

• Anthropic ($18.5 billion over 3 rounds)

• xAI ($5.3 billion from SpaceX)

• Anduril ($2.5 billion)

• Thinking Machines Lab ($2 billion)

• Safe Superintelligence ($2 billion)

• Anysphere ($900 million).

  

  A deep dive into Rubin CPX, “Another Giant Leap” optimized for long-context pipelines, reduces wasted HBM during compute‑bound prefill phases, lowers BOM/TCO, enables disaggregated prefill/decode pipelines, and will force rivals to redesign roadmaps and consider their own prefill‑specialized chips. - SemiAnalysis

A Higher Education Policy Institute report finds almost 92% of university students now use it vs. 66% the previous year; “Honestly, everyone is using it,” says Magan Chin, a master’s student, adding that “it’s evolved… now, it’s more like a study partner” and “People just call it ‘Chat’.” - the Guardian

A 2024 survey found as many as one in three high‑school seniors used AI for their 2023–24 application essays, so many colleges now invite or accept video essays. InitialView CEO Terry Crawford has signed on more than 30 schools and expects to handle at least 50,000 video submissions this cycle. - Washington Monthly

Education report on AI ethics under scrutiny for fake sources, possible AI misuse - TechSpot

A study found 52 children showed fewer physiological signs of reading-aloud anxiety — steadier voices, calmer heart rates and cooler facial temperatures — when reading to a social robot named Misty versus a human adult. - CNET

Part of school's function is to socialize kids, but hey, they will also need to be socialized to deal with robots for everything everywhere.

Everything you always wanted to know about AI but were afraid to ask - The New York Times

AI-powered religious chatbots are growing in use: Bible Chat has been downloaded more than 30 million times, Hallow reached No. 1 in Apple’s App Store last year, and some services even purport to let users "chat with God," though most aim to point users to doctrine and scripture. - The New York Times

ChatGPT can generate full PowerPoint decks with outlines, detailed slide content and speaker notes, from notes, docs, or a prompt. Using its data analysis features, can compile and provide a downloadable PPT file. - BGR

not that new and last time I tried it, didn’t work great, turned to Gamma.ai

The AI-Scraping Free Ride Is Coming to an End - New York Magazine

China says Nvidia might have breached conditions it put on Mellanox acquisition, launches antitrust probe. - The Register

Hangzhou, home of Alibaba and DeepSeek, becomes a top global AI hub. - The Wall Street Journal

Zhejiang has been a hotbed of bottom-up market economics, unlike more SOE-oriented Shanghai and foreign JV-oriented Guangzhou.

AI engineers are being deployed as consultants and getting billed at $900 per hour - Fortune

Goldman Sachs bankers explore limits of AI: ‘The risk is over-reliance’ - Financial Times

AI Will Not Make You Rich? Argues it's more sustaining than disruptive. - Colossus

I mostly agree e.g. Google and Microsoft should be able to ride AI to greater heights, it's not inherently an existential threat, but also it's a huge wave and hard enough to adapt, that tiny nimble AI-natives can punch above their weight and disrupt big players if they don't adapt (or buy them).

Vibe coding has turned senior devs into ‘AI babysitters,’ but they say it’s worth it. “I handed it off like the copilot was an employee. It isn’t,” calling coding copilots “like giving a coffee pot to a smart six-year-old.” - TechCrunch

OpenAI's Codex might be doing OK, some people are mad at Claude, it's open source so will help democratize coding agents. Certainly if you want to run in-house and point to on-prem models it might be a godsend. - Hacker Noon

AI whisperers writing specs might be the new 10x devs - Hacker Noon

Philosophers and a psychiatrist warn that outsourcing creative and intellectual effort to generative AI can diminish personal meaning, pride and skill development even as it boosts productivity. - Medical Xpress

For CIOs deploying agents: 1. set guardrails but 2. don't trust them and implement fail-safes 3. make sure your incident response plan us up to date and ready - CIO

Reuters and Harvard researcher Fred Heiding used several major AI chatbots to draft and plan a simulated phishing campaign targeting older adults, found chatbots' safety measures inconsistent and often easy to circumvent. - Reuters

North Korean spooks whip up fake South Korean military ID with help from ChatGPT, deployed in emails about ID issuance (maybe to obtain more legit IDs or info from the source?) - The Register

How robotaxi companies tell first responders to deal with them. - The Washington Post

AI-controlled drone swarms set to transform combat on battlefield - Financial Times

As Drones Swarm Battlefields, Militaries Seek Cheaper Defenses - Bloomberg

Musk’s Grok AI bot falsely suggests police lied about footage of far-right rally in London - the Guardian

Elon Musk shared an obviously AI-generated image of the Unite the Kingdom rally in London. The Arc de Triomphe is in Paris, not London. And once you zoom in,the Union Jacks and the crowd have other errors typical of AI images. - BlueSky

incredible that people are writing long, brow-furrowing posts about "blueskyism" while the owner of x, the everything app whips up one hate-filled mob after the other

According to social media, Charlie Kirk shooter was a leftist who did it for the trans, or a groyper who thought Charlie Kirk was a sellout who wasn't extreme-right and violent enough? people gonna see what they want to see, or really, just see one of those stories. FBI and Trump gonna fan the flames, not even gonna try to set the record straight.

The best artificial intelligence design of 2025 - Fast Company

MarqVision, an LA-based AI startup founded in 2021 to fight counterfeiting using computer vision, closed a $48M Series B. In China you can bring a copy of Vogue and get a tailor or seamstress to copy anything, I guess AI makes it even easier. - TechCrunch

Can AI lift VC out of its consumer tech slump? I think AI might revolutionize marketing direct to consumer brands but incorporating AI in existing categories is always hard and hardware is really hard. the guys making espresso with AI seem like a natural decentespresso.com - Business Insider

Today's VC/CEO porn courtesy of BI and Menlo - Business Insider

Follow the latest AI headlines via SkynetAndChill.com on Bluesky