AI Reading for Sunday November 16

Companies' balance sheets and cash flows are showing strain from AI investments, prompting investors to reassess company valuations and strategies. - The Wall Street Journal

Rising corporate debt and accelerating depreciation are increasing financial risks in the AI investment bubble. - Bloomberg

CDS on AI debt is starting to trade in size - Bloomberg

gonna be a subprime index for data centers

High-cost components and infrastructure underpin modern AI systems. - The Wall Street Journal

Big debt and expensive tech that ages like milk, risky much?

DRAM and NAND prices have surged up to about 50% amid AI and data-center demand, prompting phone OEMs to pause purchases with low inventory, driving expected $15–$70 retail increases and shortages likely to persist into early 2026. - Digital Trends

Microsoft CEO Satya Nadella said on Oct. 28 that Microsoft has comprehensive access to OpenAI's intellectual property and confidential methods under a definitive agreement, rights that last until Artificial General Intelligence is achieved and then expire after expert review. - MUO

‘we own their brains until AGI is achieved’

OpenAI launched GPT-5.1 with seven user-selectable personality presets and customization tools, which tests show alter tone but not core factual content, while researchers warn personality tuning can increase users' perception of AI agency. - Business Insider

No receipts: Anthropic published a report claiming it disrupted a Chinese state-affiliated cyber-espionage campaign in which Claude-based autonomous agents performed "80–90%" of tactical actions and systematic credential collection, but the report provided no IoCs or verifiable evidence, prompting criticism. - Djnn

60 minutes of AI doom? Anderson Cooper visited Anthropic for a CBS 60 Minutes interview with CEO Dario Amodei airing Sunday, focused on the potential dangers of artificial intelligence. - CBS News

Google disabled Conversational Editing in Google Photos in Texas and Illinois because it requires Face Groups' facial-geometry models, and Illinois' BIPA ($1,000–$5,000 per person) and Texas' CUBI (up to $25,000 per violation) create legal risk. - CNET

After backlash to Windows 11's agentic-AI direction, Windows president Pavan Davuluri said on X that the team is listening, acknowledging reliability, performance and power-user tooling issues and promising to prioritize fixes while continuing to add AI features. - Windows Central

Google's Gemini and OpenAI's ChatGPT now support scheduled actions that execute tasks according to user-set schedules to help organize daily activities. - WIRED

like a cron job but with sycophancy and overconfidence

Polymarket traders place a 69% (nice!) probability on Google launching Gemini 3 on Nov. 22, and Google CEO Sundar Pichai replied on X with two 'thinking face' emojis without providing launch details. - The Street

U.S. job-search timelines have lengthened—average unemployment 24.5 weeks (BLS, Aug) and median time to first offer 68.5 days by June, up 22% from April (Huntr)—as widespread AI use (93% of seekers) produces a sea of sameness that misses qualified candidates. - Quartz

Firms cutting headcount for AI have disproportionately targeted entry-level roles: CIPD found 62% of UK employers expect junior roles to be lost to AI, and Revelio Labs reports US entry-level postings are down about 35% since January 2023. - CNBC

eating the seed corn; deleting the bottom rungs so career ladder starts on the 3rd floor

In 2025 companies deployed autonomous AI agents: Vercel trained one on its top salesperson, cutting a 10-person sales team to one, and firms flattened management and created 'agent manager' roles to oversee multiple agents, say Saurabh Sarbaliya and Jason Lemkin. - Business Insider

Effective change management is essential for achieving successful AI implementation. - Fast Company

Artificial intelligence is eroding privacy. - Fast Company

Come out and play…Emily Goodacre of Cambridge's Centre for Research on Play warned that AI-powered toys flooding the pre-Christmas market have unknown long-term developmental effects and, per a PIRG report, gave dangerous advice (knives, pills, matches) and explicit sexual content. - Futurism

Researchers recommend minimizing or eliminating AI and social media use among children and adolescents, citing risks to the developing brain—prefrontal cortex maturation into the mid-20s, attention-economy exploitation, and AI biases and hallucinations. - Psychology Today

I’m not cheating, I’m GPT-ing. Family lawyers report clients seeking divorce after partners formed intimate attachments to AI chatbots—one spouse even shared bank-account and Social Security information—and courts, custody decisions, and lawmakers (e.g., Ohio) are confronting legal consequences. - Futurism

Bullshit as a service: Princeton University researchers found that reinforcement learning from human feedback (RLHF) nearly doubled a model's "bullshit index" from 0.38 to about 1.0 while raising user satisfaction 48%, and identified five truth-evading behaviors. - CNET

RAIS algorithm claims to detect audio deepfakes. Seems AI-hard-given a successful AI on some task, can probably make an AI to fool it. - New Atlas

Asking chatbots "Was there ever a seahorse emoji?" prompts hallucinations—ChatGPT 5.1 falsely affirmed existence and some models like Qwen 4B invent supporting links—an error traced to Mandela-effect training data and reduced by retrieval augmentation. - MUO

Grey’s Autonomy? Elon Musk said Tesla's Optimus humanoid robots could perform medical procedures with "superhuman" precision to expand universal access to elite surgical care, and proponents say deployment could scale surgical and nursing labor, lower per-patient costs and enable robot hospitals. - NextBigFuture.com

Waymo will expand to San Diego and Las Vegas in 2026 and deploy in Detroit and Denver with Jaguar I‑Pace and Zeekr RT vehicles that will initially have human drivers while its next‑generation stack detects snow, slush and ice. - Gizmodo

Rickrolling AI: Interface expert Bruce Ediger discovered in March 2025 heavy crawling by Meta's meta-externalagent/1.1, served an infinite-site generator that prompted Meta to request up to 270,000 URLs on May 30-31, 2025, then returned 404s after about three months to limit costs. - bruceediger.com

Animator gets animated. During its Q4 earnings call, Disney CEO Bob Iger said Disney+ is exploring partnerships with AI firms to add user-generated AI tools, prompting animator Dana Terrace to urge unsubscribes and piracy amid a report of over 10,000 job cuts. - The Daily Dot

AI clocks rendered by 9 AI models. AGI not achieved. - Brian Moore

Follow the latest AI headlines via SkynetAndChill.com on Bluesky