AI Reading for Friday August 15

Publishers reported up to 25% referral traffic declines linked to Google’s AI Overviews, though Google disputed third-party measurements of aggregate impact. - Digiday

‘Cheapfake’ AI Celeb Videos Are Rage-Baiting People on YouTube - WIRED

Introducing Gemma 3 270M: The compact model for hyper-efficient AI - Google Developers Blog

Can use it to fine-tune ‘cog’ AI for smaller data extraction and classification tasks.

Cognition Cinches About $500 Million to Advance AI Code-Generation Business After Windsurf Acquisition - The Wall Street Journal

Bolt app creation levels up, offer end-to-end coding and deployment - Business Insider

App creation platform Anything launches after raising about $11m - The Register

LLMs are good at the language part of generating and updating code but "cannot maintain clear mental models", OK for one-shotting simple tasks. - Zed

As advanced LLMs reason for longer and do more work, are you willing to pay $100k a year per developer on AI? - The Register

China hawks threaten to quit over AI chip deal, not because it's outside constitutional authority and there is no legal basis for it, but because of the yellow peril. "Superintelligence will give those who wield it the power to crush opposition, dissent, and lock in their grand plan for humanity." - Vox

There’s one small problem with Trump’s export deal with Nvidia and AMD: The Constitution says it’s illegal - Fortune

AI experts return from China stunned: The U.S. grid is so weak, the race may already be over - Fortune

Strong clickbait game

Xpeng SUV cheaper, better than Tesla Model Y? - CarNewsChina.com

Anthropic brings Claude's learning mode to regular users and devs - Engadget

Sam Altman: Never mind the launch mess — full speed ahead - Axios

What Musk, Altman and Others Say About AI-Funded ‘Universal Basic Income’ - The Wall Street Journal

Meta's standalone AI chatbot app is not very good - Bloomberg

TikTok tightened rules on AI-generated or edited content that could mislead about “matters of public importance” (“Even with labels, some edited or AI-generated content can still be harmful”), created a unified “Regulated goods and services” policy, says AI will take stuff down automatically. - Social Media Today

Entry-level tech hiring has halved at major firms since 2019 as AI automates junior tasks, squeezing new graduates and reducing internship conversions. - Fortune

Your boss wants to know how often you use AI - Business Insider

AI is moving into the apartment market, taking over work orders, lease renewals, showings and more - CNBC

An academic warned the AI boom may outstrip the dot-com bust in investor pain, citing sharp valuation swings such as CoreWeave’s decline. - Business Insider

Workday acquires Flowise low-code drag-and-drop agent builder to boost AI-powered workflows - SiliconANGLE

Startup led by Parag Agrawal launches research platform for AI agents, sounds like it competes with Tavily, Exa - SiliconANGLE

Using generative AI, researchers design compounds that can kill drug-resistant bacteria - MIT News | Massachusetts Institute of Technology

Mollick: as I predicted (and worried about) AI “personality“ is going to be the battleground for a lot of consumer Ai development. That appears to be the angle so for Grok, and the lesson OpenAI took from the backlash against retiring 4o. It may be consequential. - One Useful Thing

SweetREX, a DOGE-affiliated AI tool, is being launched to automate review of federal regulations and flag rules it identifies as candidates for elimination. - WIRED

What Does Palantir Actually Do? - WIRED

xAI was dropped from a U.S. government AI vendor list after the MechaHitler incident and reports of Grok generating antisemitic content, according to internal emails. - WIRED

AI risks in finance, wrongly flagging transactions, increasing herding and financial instability - mint

Is an AI plushie that talks to a small child a good idea? - The New York Times

Even if it was academically validated as a language learning tool I would be concerned about weird socialization and core memories bonding them to future plushie T-800s

AI mental‑health apps evade HIPAA by branding themselves as “coaching” or “wellness” tools, so users’ conversations are stored on servers and, as the piece warns, “can be extracted, packaged, and released, with or without your consent.” - Hacker Noon

The first-ever World Humanoid Robot Sports Games opened in Beijing, “Here’s hoping we can create a world where robots support human innovation instead of, well, taking over and murdering us.” - Fast Company

How to use the in-app Zoom AI tool to capture your meeting's message easily and concisely. - CNET

Follow the latest AI headlines via SkynetAndChill.com on Bluesky