AI Reading for Wednesday December 24

The White House has embraced A.I., issuing executive orders to loosen regulations while major tech hyperscalers issued over $121 billion in debt this year to finance AI infrastructure, driving record gains (e.g., Nvidia) amid warnings of bubble and job-displacement risks. - The New York Times

Money talks bigly: Meta, Alphabet, Microsoft, OpenAI and other AI firms spent $50 million on lobbying in the first nine months of 2025, while a pro‑AI super PAC is raising a $100 million war chest for 2026 and companies expand D.C. presence. - Quartz

Generally accepted AI puffery: AI development and implementation costs with accounting gimmicks can create an 'black box' that complicates financial transparency. - The Wall Street Journal

Open a chat to close the books: More than a dozen CFOs predict that in 2026 AI—including agentic, generative and predictive analytics—will move from pilots to enterprise-wide embedded systems in ERPs and finance workflows, automating close, forecasting and accounts payable while providing real-time insights. - Yahoo

Italy ordered Meta to allow rival AI chatbots to operate within WhatsApp. - The Wall Street Journal

The Pentagon and A.I. Giants Share a Weakness. Both Need China’s Batteries, Badly. - The New York Times

AI brings back coal and dirty ‘peaker’ power - CNA

AI thirst traps: Andrew Marley said in 'Hard Fork' interview that A.I. data centers use water‑intensive cooling for training and inference, prompting operators to pursue low‑water cooling, recycling and designs to reduce municipal water demand. - The New York Times

An analyst predicts that by July 2026 a US tech company will announce an AI facility abroad citing power constraints, amid data showing China added more generation 2010–2024 than the rest of the world combined and cheaper power (~$0.03 vs $0.09/kWh). - WorldNetDaily

Imperial civil service exam or Kējǔ for robots: regulations require AI chatbots to pass monthly-updated 2,000-prompt tests and refuse at least 95% of prompts about subversion or discrimination, forcing Apple to use a China-specific model such as Alibaba's Qwen3. - 9to5Mac

China's AI regulations require chatbots to pass a 2,000-question ideological test, prompting specialized agencies to emerge to help AI companies pass. - The Wall Street Journal

maybe AIs and leaders should have to pass a civics and basic decency test. seems like there is a role for democracy watchdogs to play in rating AIs for adherence to democratic and human values. people like Musk and Ellison are not subtle that they want to control it to impose their own ideology and values.

also a loyalty and ideology test seems like good reason to reject Chinese AI models even when they are open-source models hosted by e.g. Bedrock, Vertex, or Azure

Sources site inconsistencies in Megaspeed's chip inventory, as the US probes the Singapore AI startup over Chinese ownership and Nvidia chip smuggling - Bloomberg

Alibaba is reportedly planning to order 40,000-50,000 AMD MI308 AI chips, each offering up to 192GB HBM3 and priced around $12,000 per unit, creating competition with NVIDIA's H200 shipments to China. - techdeals.net

Analysts say agentic commerce could hit $1 trillion by 2030 (McKinsey) and add $115 billion to U.S. e-commerce (Morgan Stanley); Amazon debates embracing v. blocking agents, hired a corporate-development lead, blocked 47 bots and sued Perplexity while building its own agents. - CNBC

The rise of AI agents is shifting data collection from public internet scraping toward accessing more private personal data, meaning large AI companies may gather more sensitive information than before. - WIRED

Can an agent let you touch simulated grass? World models, AI systems trained to simulate environment dynamics rather than predict text, gained prominence in 2024–25 with examples like OpenAI's Sora and DeepMind's Genie, and are pursued for robotics, autonomous vehicles and medicine despite high compute and data costs. - CNET

Context matters as much as content in determining whether text is machine-generated, so people cannot reliably tell when something was written by AI. - Financial Times

everything is vibes now

Google has promised it is working on a “fix” for spam sites in its Discover app after more fake news stories topped the rankings in the UK - Press Gazette

Google's AI-generated summary wrongly identified Cape Breton fiddler Ashley MacIsaac as a sex offender, prompting Sipekne'katik First Nation to cancel his Dec. 19 concert and leading Google Canada to apologize while MacIsaac weighs legal action. - CBC

In the future, everyone is famous as a sex offender for 15 minutes

Waymo's postmortem on the San Francisco power outage says its vehicles successfully traversed more than 7,000 dark signals, used a 'fleet response' human-feedback protocol that led to backlogs and congestion. Says they don't use tele-operation, more like sending the AI remote suggestions. - Gizmodo

After pausing service during a Dec. 20 San Francisco blackout that cut power to about 130,000 customers, Waymo said it will deploy fleet-wide software updates, improved emergency-response protocols and first-responder training. - CNBC

here’s a suggestion, doen’t freeze in the middle of an intersection

AlphaFold, a Nobel Prize–winning DeepMind project, transformed biology and chemistry and remains actively evolving five years after its debut, DeepMind researcher Pushmeet Kohli told WIRED. - WIRED

More startups said to choosing Claude - Inc.

For someone who says money won't matter with AI, Elon Musk seems pretty insistent on his $56b/potential $1t pay package. - Yahoo

US bans foreign drones. - Little Flying Robots

Catfished by Clippy: recruiter spent a 30-minute interview with an AI avatar posing as a job candidate, realizing about 10 minutes in the interviewee was not human, prompting stronger identity-verification and AI-detection in hiring. - Cheezburger

I had a weird dream where Kevin Hassett was appointed Fed Chair, the Dow declined 25%, the techbros got together in Sun Valley, used the drone startups to decapitate the government, installed themselves in a coup, rounded up all the AI skeptics.

It's a world where my high school computer lab teacher Thierry Breton gets treated like a terrorist so why not.

Merry Christmas from Chiron Beta Prime, many thanks to our robot overlords, I mean protectors - YouTube

Follow the latest AI headlines via SkynetAndChill.com on Bluesky