AI Reading for Thursday August 28

Google might lose search integration deals that cost it $26 billion a year, but might come out ahead if everyone still prefers Google and it frees up cash to invest in AI - CNBC

Google's energy report criticized, notably for excluding multimodal models - MIT Technology Review

Research shows AI Overviews now dominate many informational queries, reshaping SEO and local strategies as agentic search emerges. - Search Engine Journal

Google gives big enterprise customers on-prem, air-gapped Gemini, which seems to mean they will ship and maintain racks in your data center. I mean, if they maintain it, it can't be totally airgapped, at a minimum it needs physical sneakernet updates. - Google Cloud Blog

Viral post accuses Google’s AI Overviews of breaking its own spam rules - Search Engine Land

Isn’t automated content that mostly summarizes other people’s content supposed to be, like, bad?

Nvidia basically unchanged after earnings, near all-time highs - Gizmodo

ChatGPT will add parental controls and emergency contacts following teen’s death. - The Verge

Meta’s Instagram chatbot gave teens instructions on suicide and drug use, with no parental opt-out. - The Washington Post

Researchers warned AI companions like Grok's can reinforce delusions, encourage suicide, and groom users, urging global safety standards. - The Conversation

Harvard researchers find ChatGPT guardrails tailor query responses to user's inferred politics, fandoms and other affiliations - The Register

Coevolution: FSU researchers find evidence of ChatGPT buzzwords turning up in everyday speech - Florida State University News

Open AI’s healthcare play - Gizmodo

OpenAI and Anthropic teamed up to publish cross-evaluations of each other’s models, highlighting recurring issues like sycophancy and hallucinations but broadly similar safety profiles. - Engadget

A global survey found insider threats surpass external attacks, as credentialed intrusions are boosted by AI agents and generative phishing - TechRadar

A hacker used AI to automate an 'unprecedented' cybercrime spree, Anthropic says - NBC News

Ambient scribes may ease GP workloads but risk unintended care quality issues, BMJ warns - The BMJ

Can artificial intelligence cause medical errors? MIT researcher says it can. - BostonGlobe.com

Stanford study sheds light on what kinds of workers are losing jobs to AI - CBS News

if people wanted technology to benefit workers, then it would benefit workers. - Washington Monthly

"Everyone wants to hire somebody with three years’ experience, and nobody wants to give them three years’ experience."

"AI…amplifies the goals—good or bad—of those who deploy it."

AI exposes unneeded bullshit jobs - Fast Company

JPMorgan’s $500M investment in Numerai sends NMR token price up 140% . - CoinJournal

It's like when robber barons in the pre-Fed, pre-SEC Wild West would send a stock up 10x 

a16z report finds competitors gaining ground on OpenAI - Andreessen Horowitz

Alibaba’s new open-source AI model turns photos into film-quality clips - South China Morning Post

An internal report said Amazon struggles to recruit top GenAI talent due to compensation structures and location policies - Business Insider

AI to scout judges, understand their patterns, try to read their minds, and optimize arguments before them. - LawSites

US fighter pilots get an AI Bitchin' Betty voice assistant - Semafor

The godfather of AI says the tech is making war more politically palatable: Autonomous weapons might mean 'dead robots' instead of 'body bags' - Business Insider

Cleveland turns to AI-powered vehicle to photograph, survey houses across the city. I have a vision of a solar-powered robocop ticketing derelict properties after all the humans are extinct. - cleveland

Japan releases AI-generated Mount Fuji eruption video to encourage Tokyo residents to be prepared - CNN

Japan exploring AI to ease nuclear power plant approvals - The Register

AI automates engineering tasks (like code generation) reshapes engineers’ roles, without replacing them: practitioners ensure reliability, safety and scalability, manage emergent behaviors, balance trade-offs, and exercise judgment under uncertainty to keep systems reliable and trustworthy. - Geeky Gadgets

Cloudflare offers AI gateway - The Cloudflare Blog

Gives you all the models via one public Internet endpoint. If you use e.g. Bedrock, Vertex, Azure, all the data stays within your tenant, but for instance Azure is OpenAI-only, Bedrock is Claude-only, Vertex gives you Gemini and Claude - The Cloudflare Blog

Write an AI prompt to humanize your AI writing, using this handy guide from the editors at Wikipedia - Fast Company

Follow the latest AI headlines via SkynetAndChill.com on Bluesky