AI Reading for Tuesday August 19

Anthropic raising at $500b valuation - Business Insider

A chart created by OpenAI, haven’t really checked it, Anthropic fell a bit behind in the last one but caught up a bit in the latest one to around 1/3.

for a $500b company, and considering the length of the wait, it is surprising how shambolic OpenAI’s latest release was, I am still seeing stuff moving around in the UI.

but it is pretty cool to give it a prompt and have it create this.

The progress of OpenAI models has been spectacular. - OpenAI

a bit defensive about slower progress in GPT-5 mayhaps?

OpenAI is expected to implement advertising to support ChatGPT costs but says subscriptions remain the focus. - nextpit

the enshittification of AI will be equally spectacular

Patients bring ChatGPT into doctors' offices - Medscape

What My Daughter Told ChatGPT Before She Took Her Life. Should AI be a sort of mandated reporter when some people say some things? - The New York Times

AI is a mass delusion - The Atlantic

"It's like a PhD in every field!"

"Can it reliably do college-level work, such as accurately identifying and summarizing pertinent information?"

"No."

"So what is really does is sound like an expert to people who aren't experts in that field?"

"Yeah! That's what everyone wants, right?"

SoftBank is investing $2 billion in Intel as part of its AI bet. - Business Insider

With ARM (still mostly SoftBank-owned) going into its own chips, seems like a good fit, Intel has good fab process but the third-party service situation is a disaster and they need a lot of money to reboot it.

Meta might release a $800 smart glass product in September, Apple continuous to slow-walk Apple Vision Pro with high prices, gradual improvements, slow drip of immersive video content releases. - Bloomberg

A small MIT study found lower neural engagement among students using ChatGPT, raising debate about learning impacts. - The Next Web

China is rolling out comprehensive K-12 AI curricula, teacher training, textbooks and industry partnerships nationwide. - www.chinadaily.com.cn

I asked an AI tutor agent to play the role of me, an Oxford lecturer on media and AI, and teach me a personal master’s course, based entirely on my own work. - The Conversation

everybody just gonna be talking to themselves

Duolingo CEO backtracks a bit on controversial AI-first memo - Fortune

Microsoft adds an =COPILOT function to Excel - TechRadar

Google adds AI agent data engineers (sponcon) - The Register

AI to build automated accountants - Business Insider

A survey found that while AI tools (used daily by over 75% of workers) make managers’ emails look more polished, medium- to high-level AI assistance erodes employee trust; employees often interpret heavy AI users sending more polished emails as insincere, lazy, or less competent. - SciTech Daily

Around 17% of Legal Jobs At AI Risk: Goldman Sachs analysis - Artificial Lawyer

Vibe coding: deploy faster, accrue more technical debt. - MIT Sloan Management Review

Social scientists warn AI-generated survey responses are polluting crowdsourced behavioral studies. - New Scientist

Short chats with politically tuned chatbots shifted participants’ opinions, underscoring persuasion and manipulation risks. - Computerworld

Studies show AI-edited images and AI-generated videos increase false memories and degrade visual evidence trust. - The Phoblographer

Faith-tech platforms are using AI and biometric tools to score and microtarget congregants. Use predictive analytics to A/B test prayers and see which ones work better! - MIT Technology Review

Texas attorney general accuses Meta, Character.AI of misleading kids with mental health claims - TechCrunch

The lottery ticket hypothesis reveals small subnetworks in large models can match full-network performance, explaining double-descent scaling behavior - Nearly Right

Extracting unhinged system prompts from Grok - 404 Media

‘What Jensen Huang didn’t say was that “a puppeteer behind the scenes” was controlling the robot, according to Disney, which developed it. The display spotlighted an open secret among those working on robotics: Many robots are not as autonomous as they seem.’ - The Information

The Athletes at China’s Robot Games Fell Down a Lot. Those videos of robots doing complex things fluidly seem very cherry-picked. That video of the Unitree robot in a NYC store looked stilted. Ubiquitous Rosey The Robots are quite a few years out, with a long path like robocars. - The New York Times

We made a video of Daryl Anselmo, a longtime video-game artist who now makes (mostly creepy) AI videos full-time, breaking down his workflow. - Washington Post

Follow the latest AI headlines via SkynetAndChill.com on Bluesky