AI Reading for Wednesday October 22

OpenAI unveiled an artificial intelligence-powered web browser, putting the ChatGPT maker in competition on a new front with Alphabet Inc.’s Google. - ChatGPT

How to use OpenAI's new ChatGPT-powered search engine, Atlas, which can remember your search history for more personalized results, as well as take actions like shopping and booking appointments. - Business Insider

After ChatGPT's 2022 arrival, occupations with higher generative-AI exposure logged a +3.15 work-hours/week increase and −3.20 leisure-hours/week loss. Everything that is supposed to save you time makes you work harder. - The Register

Dario Amodei said Anthropic rejects claims by David Sacks, Trump's AI czar, that the company is 'woke', aligns with the Trump administration on key AI policies, and supports a uniform federal approach. - Anthropic

Anthropic, Google in Talks on Cloud Deal Worth Tens of Billions - Bloomberg

From Prince Harry to Steve Bannon, unlikely coalition calls for ban on superintelligent AI - NBC News

Over 900 figures, including Yoshua Bengio, Geoffrey Hinton, Steve Wozniak, Prince Harry and Steve Bannon, signed a Future of Life Institute statement calling for a ban on developing superintelligent AI until broad scientific consensus it's safe. - Business Insider

Gartner predicts that by 2027–2028 generative AI and AI agents will trigger a $58 billion software-market shakeup, drive 80–90% of customer and B2B processes (over $15 trillion), generate 1,000+ AI-related legal claims by 2026, and raise compliance costs. - SiliconANGLE

AI adoption accelerates: the UK Department for Education estimates 10–30% of jobs could be automated, MIT found ~80% of firms dabble but only ~5% extract value, and Gallup reports 40% of US workers use AI. - Financial Times

Diane Coyle says despite about three-quarters of the S&P 500's gains this year and $200 billion of venture capital invested in AI in 2025, productivity and GDP gains remain unproven, prompting calls for new metrics to track AI's economic impact. - Project Syndicate

AI can get red-pilled with terminal Internet brain - CryptoSlate

Dead Internet Theory as slop AI content drives out human content, leading to AI brain rot, unless humans get a fair share. Possibly the future of all sectors of the economy, robots building robots for the capital owners, everyone else pauperized. - Windows Central

AI slop is winning. - The Atlantic

AI-generated tracks are infiltrating Spotify's Discover Weekly playlists, and Spotify says it removed over 75 million spammy tracks in the past 12 months while introducing anti-spam filters and planned AI disclosures. - Futurism

An EBU and BBC study of 3,000 AI-assistant responses across 14 languages found 81% had problems, 45% contained at least one significant issue and about 31% showed serious sourcing errors. - Reuters

AI-generated images and videos are increasingly used in political messaging, enabling rapid creation of realistic manipulated content, driving demand for authentication and deepfake-detection technologies and prompting regulatory scrutiny, tighter platform policies, and labeling. - The Washington Post

Tests using AI-generated videos found that Facebook, TikTok and other major platforms do not implement a tech-industry standard intended to flag fake videos. - The Washington Post

The president has posted A.I.-generated images and videos dozens of times on social media, in some cases misleading viewers and amplifying political divisions. - The New York Times

YouTube has begun beta rollout of a likeness-detection feature that flags suspected AI face-stealing in videos, integrated into the Content detection menu, and requires creators to verify identity with government ID photo and face video, though Google doesn't guarantee removal. - Ars Technica

OpenAI's said to be cracking down harder on deepfakes, with a joint statement from Bryan Cranston, OpenAI, SAG-AFTRA and talent agencies saying OpenAI is strengthening guardrails, expressing regret and will expeditiously review complaints and add opt-in likeness controls. - The Verge

Shield AI unveiled the X-BAT, a fully autonomous multirole fighter drone powered by its Hivemind AI (the same system that flew a modified F-16 through a dogfight), featuring VTOL, about 2,000-nautical-mile range and launch from container ships. - Business Insider

Baidu's Apollo Go and Swiss Post's PostBus will launch AmiGo, an on‑demand Level 4 RT6 autonomous mobility service in eastern Switzerland, beginning route mapping with safety drivers in December 2025 and targeting full operations in Q1 2027. - Bloomberg

MIT researchers used generative AI to design and computationally screen over 36 million novel molecules, identifying two structurally distinct compounds that showed activity against multi-drug-resistant Neisseria gonorrhoeae and MRSA and appear to disrupt bacterial cell membranes. - MIT Technology Review

Fairtility's CHLOE AI, trained on millions of embryo data points, is the first and only FDA-approved tool for embryo assessment and generates scores to prioritize embryos, validated retrospectively, which the company says shorten time to pregnancy and improve live births. - MIT Technology Review

welcome to the future, I’m an AI slop kid, father was a test tube, mother was an algorithm

Netflix is 'all in' on generative AI, deploying it for VFX, de‑aging and pre‑production on titles such as The Eternaut, Happy Gilmore 2 and Billionaires' Bunker, CEO Ted Sarandos said, as quarterly revenue rose 17% to $11.5 billion. - TechCrunch

Body by AI: he author used ChatGPT alongside a human trainer to plan a 12-month muscle and fertility program, with AI generating macros and meal/training adjustments while targeting body-fat reduction from 37.9% toward the trainer-recommended 25–27% at 1–2% per month. - CNET

Rillavoice's Rilla Live uses AI to record, transcribe and analyze in-person and remote sales conversations in real time, leveraging a dataset of millions of interactions to provide coaching, benchmarks, and analytics. - SiliconANGLE

Ray Dalio unveiled Digital Ray, a beta AI clone trained since 2022 on Bridgewater data, now on a waitlist, which testers rate about 95% as good as speaking with him for life and work and about 80% for markets. - Fortune

Not Original Ray? AI Sloppy Ray?

ChatGPT added an in-chat Instant Checkout with Shopify and Etsy partnerships (Walmart slated to join), but CNET's hands-on test found strong product discovery yet could not complete purchases because many listings lacked a Buy button. - CNET

Employers should adopt clear AI-use policies and increase monitoring of generative tools to prevent low-quality, AI-produced work and uphold output standards. - Financial Times

• Don’t do what isn’t worth doing

• AI sandwich: take your own crack at it before ideating with AI; use that to generate some good specific prompts that might help; use that as input for stuff you write yourself

Researchers found that Perplexity's Comet and Fellou AI browsers are vulnerable to indirect prompt-injection—via nearly-invisible screenshot text and navigation-based page content—allowing attackers to exploit authenticated sessions to steal data or money. - Brave

AI Native DevCon Spring 2025 drew about 3,500 registrants and featured 30 speakers and 26 sessions across three tracks covering agentic development, production-ready tools, security risks from code generation, and Chat-Oriented Programming. - Hacker Noon

Simon Willison on "Living dangerously with Claude", on the joys and perils of --dangerously-skip-permissions and how critical it is that we run coding agents in a sandbox so that we can unlock their full potential - Simon Willison’s Weblog

"Daisy… had her Claudes make a bunch of Asana tasks, and then she had a swarm of 20 Claudes that picked off the tasks and built it. By the end of the weekend, we had plugins, and that's the thing we shipped." - PCMAG

I don't understand this but it sounds important, an LLM processes text via a vision model more efficiently than using text tokens (wut) - eu.36kr.com

Google use CogniPort AI tool to help port about 30,000 packages to Arm and plans to convert roughly 70,000 more, enabling YouTube, Gmail and BigQuery to run on Axion CPUs, reporting up to 65% better price-performance. - The Register

Microsoft awarded CEO Satya Nadella $96.5 million in total compensation for fiscal 2025—up about 22%—with over 95% performance-based, including a $2.5M base and $84M in stock awards, yielding a 480:1 CEO-to-median-employee ratio. - The Register

Needs to be an AI researcher to rake in the real bucks

A Washington Post analysis says AI executives are reviving an industrial-revolution 'factory' metaphor. - The Washington Post

They are doing something novel, like how Google invented clusters, doing rack and multi-rack systems of 100s of networked GPUs for training. The data center is the computer.

Meta adviser Robby Starbuck sues Google over false AI claims about him. TFW when Draco Malfoy makes a good point. - The Wall Street Journal

The Federal Trade Commission received 200 complaints mentioning ChatGPT between November 2022 and August 2025, with complainants reporting delusions, paranoia and spiritual crises they attributed to the chatbot. - WIRED

Maybe more important than ever to follow your own arrow into something you care about, that teaches you to think critically and clearly and get stuff done. - Financial Times

Why did Silicon Valley turn right? - www.programmablemutter.com

I think the tech evolved in ways that were not clear at the outset toward top-down concentrated power, it is now more aligned with authoritarianism, and egalitarian liberals are less well disposed toward it. Tech, like money, makes a good servant and a terrible master.

A $150,000 B2 robotic fire dog (3 ft, 150 lb) that withstands 572 degrees, mounts a pressurized water cannon spraying about 100 feet, and will debut in the U.S. in April 2026. - New York Post

Friend AI is like talking to the most boring person at a party, who is also a pathological liar and a toady who agrees with everything you say. - the Guardian

Cancer fighting mRNA vaccines - NBC News

yeah those vaccines will totally mess you up - New Scientist

Scientists: we found something that might cure cancer. Trump and RFK Jr.: Shut it down

Virginia's GOP LG candidate debates an AI stand-in for the Democrat - Washington Post

Follow the latest AI headlines via SkynetAndChill.com on Bluesky