AI Reading for Tuesday-Wednesday July 30-31

I got a little backlogged yesterday, did you miss me?

---

SAM 2 Segment Anything Model is multi-modal, does real-time video segmentation at 44 frames per second, previous version of open-source module worked with still image segmentation, which was no good for real-time processing. Seems like it will enable a lot of robotics stuff. - THE DECODER

Meta is still the worst though, I hate it when Yann shills for them.

---

Zuck talked to with Jensen Huang at SIGGRAPH, said he likes open source AI. We get it, you don't want Google or Apple or OpenAI or the Feds saying you can't data-rape the world. - Business Insider

---

Meta moves on from chat with a celebrity to, influencers can make their own custom chatbots. - The Verge

---

Meta settles with Texas AG for $1.4b over using facial recognition without permission. - The Texas Tribune

---

Meta prompt injection prevention model defeated by random spaces. - The Register

---

Nvidia offers dev platform for humanoid robots. - Tech Monitor

---

Apple released a paper on its responsible approach to training its AI models, appears to use Google TPUs over Nvidia for training. - Reuters

---

Apple intelligence drops in iOS, MacOS betas. - The Register

---

Saying something is AI-powered is a big turnoff for consumers. - WSU Insider

---

Movie editors face elimination by AI. - NY Times

---

Management consultants are making bank on AI. - FT

---

LLM models help robotaxis deal with long-tail scenarios. Kind of like back-seat drivers? Maybe they can just take suggestions from passengers. - The Conversation

---

Your emotional support IoT device is not weird at all. - The Verge

---

Canva buys leonardo.ai, which developed a deepfake problem. - Fortune

I'm not sure products like Sora, RunwayML, Kling, Luma etc. can sidestep a major deepfake problem. (see my Jerome Powell video)

---

Golf Digest makes an AI iPhone app that analyzes your swing. - Golf Digest

---

ABA says lawyers can't use not understanding AI as an excuse for briefs with hallucinations and bad counsel. - LawSites

---

Amazon builds solar farms with AI robots. - About Amazon

---

Baidu claims breakthrough in avoiding hallucinations. - VentureBeat

---

AI to optimize nuclear reactor designs. - Techxplore

---

AI to control nuclear reactors. - Purdue

---

Synchron integrates brain with Apple Vision Pro. - ZDNET

---

Synchron brain-computer interface adds OpenAI integration. - NewsBytes

---

Bipartisan group advances a law to make platforms liable for deepfakes. Platforms want, if you do at least X (currently nothing), you cannot be held liable. - The Verge

---

Microsoft calls for legislation on deepfake fraud and abuse. - The Verge

---

Google's multi-billion-dollar investment into Anthropic draws scrutiny from the UK competition agency. - Reuters

---

Reviews of the Apple Intelligence betas are starting to come in. - Tom's Guide

---

Imagine what Frank Abagnale Jr. could have done with AI, or the Great Imposter Ferdinand Demara (who was the imposter Frank Abagnale aspired to be) - NewsBytes

---

OpenAI begins rolling out voice assistant. - Yahoo Finance

---

Perplexity unveils revenue share. - VentureBeat

---

Tesla's robotaxi looks dubious to basically everybody. - NY Times

---

Tesla analyst says FSD claims are BS. - Bloomberg

---

The shitpost Olympics. - Platformer

---

Air Force picks five firms to build autonomous drone wingmen. - Defense News

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky