AI Reading for Thursday August 22

Silicon Valley comes out in force against AI bill. - The Atlantic

---

OpenAI says California's AI bill will hurt US innovation, competitiveness, and national security - Bloomberg

---

a16z ranks the most popular consumer AI apps. - Andreessen Horowitz

---

Napkin adds pizzazz to your presentations - CNET

---

Adobe 'Magic Fixup' helps automate complex edits. - VentureBeat

---

There’s an AI image watermark removal SaaS now, natch - HackerNoon

Ideogram has a big new update. - Tom's Guide

---

‘Feel the AGI’ types get a little butthurt when you mention ChatGPT can’t count the R’s in strawberry, kind of miss the point. - Reddit

It knows how to spell 'strawberry' or 'raspberry' just fine but it can't count the R's correctly which is something people using AI should probably understand - Druce.ai

---

How to use Claude, including artifacts and optimizing prompts in the studio. - Geeky Gadgets

Use Claude to build a web app in minutes for free - Geeky Gadgets

I kind of think Claude is better at my tech questions than GPT-4o, although lmarena.ai doesn’t agree.

The Information says Claude is popular in AWS and Llama isn't. - The Information

---

If you publish an API, you will probably get degens asking you to debug their obviously wrong AI-generated code. - Medium

---

A look at the facial recognition systems that are becoming ubiquitous. - WIRED

Cops object, say it compromises their privacy. Law enforcement sometimes argues for more surveillance, saying if you have nothing to hide, etc., but they aren't so keen on being monitored with facial recognition and body cameras. - ESPN

---

Waymo is on an offensive to grow rapidly. - TechCrunch

---

Tesla's FSD is years behind, Musk's BS notwithstanding, they would have to make it actually work first, and then test robo-cars on the streets and ramp up robo-taxi service. - Rolling Stone

Musk had to reveal who co-invested with him in Twitter, includes Saudis, Chinese, well-heeled SV funds, Fidelity, who had to mark it down to 28.5¢ on the dollar - Fortune

---

Medical AI stuff - Epic adding 100+ AI features to MyChart, automating doctor paperwork and letting patients ask if they can resume playing pickleball yet. - CNBC

AI can tell which patients are going to need joint replacement.

---

Microsoft will try again with recalled Recall feature. - Ars Technica

---

You need clean data security and governance to implement RAG Copilots, if users have some obscure access to something the AI will find it. - The Register

If it's pure retrieval-augmented generation, it should only answer questions based on data the user has access to, but if it's not properly implemented to run in userspace, you’re gonna have a bad time.

---

Reddit guy solves 633 Leetcode problems in 24 hours with AI for $9. I think AI coders can probably solve unseen easy problems decently, but if they solved hard problems well, it's probably because they trained on the answers. That might change though. - FavTutor Articles

---

Francis Ford Coppola apparently just asked ChatGPT for all the bad reviews of The Godfather and his previous films, then put made-up quotes in his trailer - Vulture

Maybe they asked an intern to do the research and they thought ChatGPT was always accurate or something. Maybe a miscommunication. I can’t believe someone would just make up quotes?

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky