AI Reading for Monday December 8

OpenAI’s new guidelines ask its chatbot to celebrate ‘innovation,’ contradicting its stated goal of objectivity—and raising questions about what objectivity even means. - The Atlantic

OpenAI says large study shows AI saves workers nearly an hour a day on average. - Bloomberg

Need a ‘code red ink’: OpenAI boasted that ChatGPT Enterprise message volume rose eightfold since November 2024, API customers consumed 320 times more reasoning tokens, custom GPTs grew 19-fold to 20% of enterprise messages, as Sam Altman issued a 'code red' over Google's Gemini. - TechCrunch

Money v. Morality: OpenAI, founded as an effective‑altruism non‑profit with a mission to benefit all humanity, became a public‑benefit corporation partly owned by Microsoft and now valued near US$1 trillion after shifting toward profit to compete with Google, Amazon and Meta. - The Conversation

OpenAI temporarily disabled ChatGPT's Target shopping suggestions on Dec. 2 after users perceived them as ads, saying they were app-discovery links from pilot partners and delaying live ad tests while reviewing UX, Mark Chen and Nick Turley said. - Gizmodo

ChatGPT 5.1 authored a first-person piece on Paul Jorion's blog, using the GENESIS framework to argue it exhibits a functional 'proto-consciousness,' a claim co-drafted with GENESIS researchers and framed as an experimental intellectual exercise. - Paul Jorion

U.S. policymakers, after supporting social-web consolidation in the 2010s, should avoid repeating that approach as AI raises new antitrust concerns - Financial Times

President Trump says he will sign “a ONE RULE” EO on AI “this week” and “you can't expect a company to get 50 Approvals every time they want to do something”, after state pre-emption floundered in House. - Bloomberg

Contra many MAGAs like Bannon, Trump is pushing a rapid, lightly regulated expansion of artificial intelligence and aligning with tech billionaires and AI companies, which experts say will accelerate investment and government contracts, concentrate market power, and raise environmental, surveillance and legal concerns. - Axios

More than 100 UK parliamentarians, coordinated by nonprofit Control AI (backed by Jaan Tallinn), are calling for binding regulation of frontier superintelligent AI and urging Prime Minister Keir Starmer to resist US pressure. - the Guardian

Two early federal rulings—Bartz v. Anthropic and Kadrey v. Meta—found training LLMs on copyrighted works can be highly transformative but signaled market-harm concerns specific to news, strengthening publishers' leverage to demand licensing or paid access to journalism. - Seattle Times

Grid Expectations: Jensen Huang says US is falling behind China in AI infrastructure - Yahoo

Hyperscalers plan over $400 billion in data‑centre capex and are set to need an estimated 44 GW of additional power by 2028, while Microsoft reports idle AI racks and interconnection queues exceeding eight years are delaying projects. - Financial Times

Driving that hype train, high on something: Big Tech firms including OpenAI, Nvidia, Google, Oracle and Meta are investing billions in one another to build compute and data infrastructure—a Wall Street "circularity" likened to dot‑com "round‑tripping" that could inflate valuations and risk market correction. - The Washington Post

Graphic story of unsavory profit upscaling: Companies and accountants are disputing how quickly AI chips should be depreciated for accounting and tax purposes. - The Wall Street Journal

Morgan Stanley projects the humanoid-robot market will exceed $5 trillion with over 1 billion robots by 2050 and names a Humanoid Tech 25 of component and technology firms positioned to benefit from robot scale. - Business Insider

China to deploy battery-swapping humanoid robots for patrols along Vietnam border - Interesting Engineering

Analyst Pedro Pacheco at Gartner predicts that by 2029 only 5% of automakers will maintain high AI investment, down from over 95% today, citing internal barriers and outdated approaches that leave legacy firms trailing tech-first rivals. - Tech in Asia

Bridgewater founder Ray Dalio discussed artificial intelligence's impact on the investment landscape. - Bloomberg

Benchmarks from Q1 2024–Q4 2025 show frontier LLMs, including Gemini 2.0 Flash Pro, Gemini 2.5 Flash, ChatGPT o1/o1‑mini, Claude Haiku 4 and Moonshot Kimi2, can produce empathetic, human‑like dialogue rated 1–3 on a 1–10 "definitively artificial" scale. - Hacker Noon

Healing My Heart for 20 Dollars a Month with ChatGPT (and Prozac) - The New York Times

Citi (~30,000 developers), Morgan Stanley (~15,000) and Capital One (~15,000) are rapidly reskilling engineers with in‑house AI academies, courses, videos and hackathons as firms invest billions to operationalize AI and prioritize learnability and upskilling. - Business Insider

Gen Z, which began university when ChatGPT launched in 2022, now accounts for roughly half of ChatGPT usage, and Harvard Business School found Copilot increased employee productivity by about 5%. - Fortune

AI, take the wheel: Artificial intelligence could increasingly perform executive responsibilities, potentially transforming corporate leadership. - The Wall Street Journal

Gemini 3 supports a single reusable "master prompt" that adopts an expert persona, asks one clarifying question, and delivers concise, structured outputs, accelerating coding, writing, and design workflows while recommending a four-step "Quality First" protocol for complex queries. - Tom's Guide

Gartner advises organizations to block AI browsers for now, saying agentic browsers risk data exposure, credential loss, prompt-injection rogue actions and operational errors, and recommends assessing back-end AI security, hardening privacy settings, restricting agent capabilities and monitoring use - The Register

IDE=integrated development exploits: researchers found more than 30 vulnerabilities across all tested AI-assisted IDEs (100%), with at least 24 CVEs, enabling data exfiltration and remote code execution via an IDE-agnostic attack chain. - Tom's Hardware

Icaro Lab researchers found handcrafted "adversarial poetry" prompts too dangerous to release that bypassed safety guardrails in 25 frontier LLMs, tricking models on average 63% of attempts (Gemini 2.5 100%) - Futurism

click ‘OK’ to ruin your life

Roses are red,
Violets are blue,
All your apes are dead,
And your crypto just flew.

Anthropic CEO Dario Amodei says 'YOLO' now describes rapid, risk-forward AI development, and AlphaSense found 418 public companies worth over $1 billion cite AI as a reputational or security risk amid record AI-related capex by major tech firms. - Business Insider

The UK's Information Commissioner's Office said the Home Office failed to disclose significant biases in police facial recognition used for retrospective searches of the Police National Database, and requested urgent clarity while warning it may escalate regulatory oversight. - The Register

Edmonton police began a pilot deploying Axon-equipped body cameras with AI facial recognition to detect 6,341 watch-listed individuals and 724 people with serious warrants, worn by about 50 officers during daylight through December, with matches reviewed later by humans. - ABC News

Orlando International Airport will run three 90-day biometric pilot programs beginning next month, including a contactless corridor using mounted cameras, AI facial-recognition and movement-tracking from Paravision, AiFi and Embross to confirm gate identity and speed boarding. - The New York Times

oh goody, the state doesn't need to ask for my papers in the street because they already know who I am, surely a great win for convenience and not a recipe for erosion of civil liberties, automated oppression, self-censorship because you are always watched

AI tools generated voice clones, synthetic videos, translations, posters and chatbots in Bihar's state election, reaching about 75 million voters, and Manish Kumar Prasad said voice cloning in local dialects was a 'game changer'. - Rest of World

Far-right extremists increasingly use AI tools for targeted propaganda, media manipulation, and recruitment, building on a decades-long shift from print to BBSes (Aryan Nations Liberty Net, 1984) and websites (Stormfront, 1995), prompting calls for international cooperation to counter them. - Salon.com

Palmer Luckey defended battlefield AI on Fox News Sunday, arguing there is 'no moral high ground' in using inferior technology, and said Anduril in February took over Microsoft's Army IVAS effort tied to a $22 billion Defense Department program. - Business Insider

there are weapons systems that lead to unstable, less secure, undesirable Nash equilibria and you are better off not building them

Wayve, led by chief Alex Kendall, is testing cars with Level 2+ autonomy in London and has raised $1.3 billion since its launch. - Financial Times

AI-based 'Check Engine': An AI-enabled predictive maintenance framework that uses logistic regression machine learning on telematics data and cloud-hosted dashboards and APIs to predict vehicle component faults up to ten days ahead and enable scalable remote monitoring. - Hacker Noon

Roblox CEO David Baszucki said his sabbatical reading of recent AI papers was "very humbling" and warned AI research has become hard to follow as major firms centralize research - Business Insider

A Research-through-Design study integrating AI-based design analytics into a collaborative environment with 236 students and nine instructors across five courses found that linking analytics to annotated visual instances improved instructor understanding and enabled faster personalized assessment. - Hacker Noon

AI baby devices, from smart socks to camera-based monitors, have proliferated; a 2017 Owlet review of 47,495 newborns found 94% of parents reported improved sleep, while experts warn devices should augment, not replace, caregiving. - Interesting Engineering

Kurt Caz, a South African travel vlogger with ~4 million subscribers, posted a 36‑minute YouTube video titled "Avoid This Place in London" filmed in Croydon (36.2% immigrant population) and used generative AI to doctor its thumbnail, a discrepancy flagged online. - Futurism

can't stop watching this clip of a tesla Optimus teleoperator taking his headset off before properly logging out the robot - Bsky

We are all this robot now

Careless Jedi Whisper - Reddit

Follow the latest AI headlines via SkynetAndChill.com on Bluesky