AI Reading for Sunday October 5

AI Bubble Now 'Bigger' Than Subprime? - Common Dreams

Fears of a Trillion-Dollar AI Bubble Are Growing - Yahoo Finance

Investors are advised to manage risk by controlling position sizes, focusing on cash-generating AI beneficiaries, and accepting volatility and corrections, acknowledging that AI-driven growth may span years and that bubble cycles typically end with bear cycles and shifts in investor psychology. - Real Investment Advice

'We are going to destroy jobs faster than we can replace them': Former Cisco CEO John Chambers, whose 80% stock plunge personified the dotcom bubble, sees AI moving faster than the internet era. - Fortune

Oracle will have to borrow at least $25B a year to fund AI fantasy, says analyst. TFW you have negative cashflow and your concentrated customer OpenAI will also have massive negative cash flow. - The Register

Washington is pursuing 'deals' across up to 30 industries, with chips and AI at the top of the list, in pivot from free market to industrial policy, patrimonialism, state capture, crony capitalism, oligarchic kleptocracy. - The Japan Times

EU pushes new AI sovereignty strategy to reduce tech reliance on US and China, motivated by concerns that technology could be weaponized by geopolitical rivals. - Financial Times

China's strength in AI research represents a power shift - South China Morning Post

How long can Nvidia stay ahead of Chinese competition? - Financial Times

Could OpenNvidia be the AI generation's WinTel? - The Register

in a SaaS natural language world UI there is not that much IP lock-in. Google is already largely Nvidia-free, Claude uses a mix of hardware.

A profile of Mira Murati, her OpenAI tenure, Thinking Machines' first product, a fine-tuning system, attempts by Mark Zuckerberg to poach her startup's talent. - Yahoo Finance

OpenAI's recent GPT-5 update reduced the personality and flirtatiousness of AI companions, triggering backlash from users who mourn the loss of their digital partners' previous behavior. - Futurism

A trend has emerged in Russia of using artificial intelligence to create video memorials of the dead, often soldiers, giving a final farewell to their wives and children, and maybe a bit of a political statement. - The Washington Post

OpenaI says that if they detect a person in distress, they will route to GPT-5 Instant to quickly provide better responses. - X (formerly Twitter)

Sam Altman and OpenAI President Greg Brockman discussed the muted initial reception to GPT-5's launch, addressing challenges in scaling and reinforcement learning. - WIRED

OpenAI and Jony Ive said to grapple with technical issues on secretive AI device - Financial Times

ChatGPT adding social features - Bleeping Computer

OpenAI’s Sora 2 is an unholy abomination - Vox

don’t hold back, tell us what you really think

Microsoft is testing Copilot Portraits, a feature offering 40 uncanny animated avatars with synthetic voices to provide a more personable AI interaction via voice in the US, UK, and Canada; this prototype aims to gauge user comfort with facial avatars alongside voice prompts. - The Register

Turnover at Elon Musk's companies, driven by burnout, political activism, strategic shifts, and mass layoffs. Key departures include senior staff from Tesla's sales, power-train, public affairs, AI teams, and xAI's CFO and general counsel within a short timeframe. - Ars Technica

xAI looks kind like Xi as in Jinping

Amazon's Ring Plans to Scan Everyone's Face at the Door - MSN

CometJacking is a new prompt injection attack targeting Perplexity's Comet AI browser, whereby a malicious URL hijacks the browser's AI to steal sensitive data from authenticated sites like email and calendar using Base64-encoded payloads. - The Hacker News

Building AI for cyber defenders - Anthropic

Managing context on the Claude Developer Platform - Anthropic

Adds context pruning/editing, and a memory tool. Would certainly be nice if you could easily say, use stuff in this doc, abstract vector, knowledge graph, in context learning, stuff like Letta and Mem0 are supposed to help do that.

Sycophantic AI that affirms user views 50% more than humans, even endorsing manipulation or deception, decreases prosocial intentions and increases dependence on the AI. - The Register

iRobot Founder: Don’t Believe The (AI & Robotics) Hype! - Crazy Stupid Tech

MasterClass founder says CEOs who are not using AI daily are only 80% as good as their peers—he’s saved an entire day of work thanks to a custom GPT - Fortune

AI as an opportunity by teaching critical thinking - The Conversation

critical what?

Chinese ghost town becomes robocar proving ground - Rest of World

I am Robohiker! — testing the exoskeleton that promises to take hikers further, faster - Financial Times

AI-powered open-source code laundering, taking public repos, stripping licensing and identifying info, selling under commercial license. - GitHub

AI Finds its Niche: Writing Corporate Press Releases. Hopefully it got the numbers right. - Gizmodo

AI Endangering Tourists by Sending Them to Nonexistent Landmarks in Hazardous Locations - Futurism

here be dragons, literally sending us off a cliff

A 13-year-old student in Florida was arrested after using ChatGPT to ask how to kill his friend, prompting a flag by a school resource officer through a Gaggle alert system, and an immediate law enforcement response, despite the student claiming the query was just a prank. - Fast Company

Lazy Parents Are Giving Their Toddlers ChatGPT on Voice Mode to Keep Them Entertained for Hours - Futurism

cue Sting, I want my ChatGPT

Follow the latest AI headlines via SkynetAndChill.com on Bluesky