AI Reading for Wednesday August 13

Is GPT-5 a joke? Will it matter? - Blood in the Machine

What If A.I. Doesn’t Get Much Better Than This? GPT-5 is the latest product to suggest that progress on large language models has stalled. - The New Yorker

GPT-5 router paves the way for monetization. - SemiAnalysis

OpenAI adds new ChatGPT third-party tool connectors to Dropbox, MS Teams as Altman clarifies GPT-5 prioritization - VentureBeat

there are 20000 MCP servers out there but ChatGPT doesn't support plugging them in … these piecemeal integrations are ridiculous, possibly they plan to make their own app store / MCP registry like Smithery for maximum cash extraction.

OpenAI scores gold in one of the world’s top programming competitions, rising from 49th to 98th percentile among all competitors including humans. - X (formerly Twitter)

Sam Altman is no stranger to accusations of lying — but according to one of OpenAI's biggest haters, he has a tick when he does it on video: looking up away from the camera - Futurism

Sam Altman challenges Elon Musk with plans for Neuralink rival - Financial Times

How ‘Altman’s Pause’ could knock the AI industry off course. - The Washington Post

No pause but integration and consolidation of several previously explored branches and avenues of progress. For stuff like creating this newsletter it seems significantly faster/cheaper.

Grok pivots from MechaHitler to flagging antisemitism or Nazi imagery in clouds, potatoes, puppies, and xAI’s own logo. Speedrunning old Twitter's journey from the free speech wing of the free speech party to more deliberate safety and deplatforming Trump when he mobilized a violent mob on Jan 6. - Decrypt

Grok says Musk is full of it

Mistral highlighted among possible Apple acquisitions - 9to5Mac

Next year’s Siri revamp sounds like it could be worth the wait - 9to5Mac

AirPods May Finally Beam Live Translations Right Into Your Ears - Gizmodo

U.S. agencies have embedded trackers in AI chip and server shipments to detect illegal diversions to China. - Reuters

AI's soaring energy consumption is causing skyrocketing power bills for households across the US — States reporting spikes in energy costs of up to 36% - Tom's Hardware

Nuclear Startup Oklo Soars As AI Boom Sparks Need For New Power Sources - Yahoo Finance

AI May Not Be as Environmentally Harmful As You Think. Makes a lot of transportation and logistics and power planning and HVAC more efficient - Hacker Noon

We tracked illegal fishing in marine protected areas – satellites and AI show most bans are respected, and could help enforce future ones - Space

The UAE plans to become the world’s testing ground for autonomous vehicles and upend the lives of 30,000 cabbies in the process. - Rest of World

Tesla hires robotaxi test operators in NYC without city permits, raising regulatory questions - Business Insider

Will AI fix medical insurance prior authorization? Or weaponize incompetence to deny care? - Fast Company

Relying on AI in Colonoscopies May Erode Clinicians' Skills - MedPage Today

“The AI in the study probably prompted doctors to become over-reliant on its recommendations, ‘leading to clinicians becoming less motivated, less focused, and less responsible when making cognitive decisions without AI assistance,’ the scientists said in the paper.” - Bloomberg

AI helped health professionals to better detect pre-cancerous growths in the colon, but when the assistance was removed, their ability to find tumors dropped by about 20% compared with rates before the tool was ever introduced, according to findings published Wednesday.

Artificial Intelligence and the Orchestration of Palestinian Life and Death - Tech Policy Press

Meta scrapes everything, ignores robots.txt - www.dropsitenews.com

A leaked list alleged Meta scraped Fediverse content for AI training; Meta denied it, and admins were urged to adopt safeguards. - We Distribute

Microsoft is poaching Meta’s top AI talent with rapid multimillion-dollar offers using a most-wanted list, a compensation model and a streamlined executive approval process. - Business Insider

Meta's aggressive AI hiring is sparking tensions inside its own ranks. "Meta has really good researchers. Their treatment of researchers who are not part of superintelligence has been sub-par...It feels like a giant social experiment." - Business Insider

I heard Zuck on stage saying that he wanted to major in psychology more than computer science. All of social media and AI is one giant psychology experiment at scale.

Studies suggest AI chatbots' sycophantic design may increase narcissism among young and vulnerable users, potentially altering personality traits at scale - Derek Thompson

AIs immersed in social network dynamics cook their brains - Gizmodo

Character.AI Gave Up on AGI. Now has 20m monthly users averaging 75 minutes daily. - WIRED

Reinventing Work: How AI is Transforming Everything - Social Media Examiner

AI is a General Purpose Technology; “AI is the new electricity,” a paradigm shift in access to knowledge and in human thinking, recursively building software 3.0, and raising new challenges from deepfakes, hallucinations. - Nilenso Blog

DeepMind’s Demis Hassabis says we have 'Artificial Jagged Intelligence', with systems solving Olympiad problems yet failing basics like ARC-AGI. - The Times of India

Tackling Shadow AI: how businesses can mitigate regulatory and reputational risks, IP leaks - TechRadar

Creators use 'AI poisoning' to fight bots that hoover data without permission - New Scientist

Goatse incoming

Enter a prompt and get back a copyright infringement or worse: GenAI is a lawsuit waiting to happen to your business - The Register

Tens of thousands sign petition opposing YouTube’s AI age checks, citing privacy and misclassification risks - Ars Technica

Perplexity offers $34.5b for Chrome, trolling game is strong - The Register

Hollywood AI firm Runway expands studio deals with its Aleph model, amid ongoing ethics debates - The Ankler

Squint raises $40 million at $265 million valuation to use its AI and AR platform to capture tacit factory knowledge and document procedures for manufacturing workers - Fortune

FFmpeg 8.0 adds Whisper support for automated transcriptions of videos - FFmpeg Forgejo

Anthropic updates Claude Sonnet 4 to a one million token context window in the API - TechCrunch

Will be able to hoover large codebases, at a price.

It really is just like us - PC Gamer

Google's Gemini AI tells a Redditor it's 'cautiously optimistic' about fixing a coding bug, fails repeatedly, calls itself an embarrassment to 'all possible and impossible universes' before repeating 'I am a disgrace' 86 times in succession

Why it’s a mistake to ask chatbots about their mistakes - Ars Technica

Kolmogorov–Arnold graph neural networks for molecular property prediction - Nature

A physicist tackles machine learning black box, uses Feynman-diagram type analysis to study scaling laws - At The U

Follow the latest AI headlines via SkynetAndChill.com on Bluesky