AI Reading for Sunday July 6

Robotaxi War in Texas - Tesla versus Waymo - NextBigFuture.com

Shocking incidents – from drivers trapped in burning vehicles to dramatic stops on the highway – have led to questions about Tesla safety. Leaked files reveal unresolved autopilot defects and design quirks like jammed door handles that hamper rescues, raising accountability questions. - the Guardian

DeepSeek R1 is still the top open source model but hasn't really challenged the top proprietary models. - SemiAnalysis

The Coder ‘Village’ at the Heart of China’s A.I. Frenzy - The New York Times

DeepSeek torched Llama and made some AI researchers rich jumping to Meta. They caught up with Baidu on censoring Tienanmen, having to deal with a bunch of politicos surely slowed them down. Possibly open-source models will just remain 6 months to a year behind? Or maybe R2 will be 6 months or less behind GPT-5 which is tipped for release this month?

Open source project makes strides in bringing CUDA to non-Nvidia GPUs with drop-in compatibility and high performance - TechSpot

Big if true. But Nvidia is very strategic, moving up the value chain and integrating the whole datacenter-as-a-computer. You have to consider the total cost of training or inference. Nvidia can essentially capture the lifetime TCO cost difference between it and the closest competitor. On some tasks like inference and simpler pytorch jobs that don’t need a datacenter competitors will catch up.

Oh cool, DSPy is its own framework, also, if you have good evals you can make a system that constantly self-optimizes its own prompts as people use it. - MarkTechPost

AI might now be as good as humans at detecting emotion, political leaning and sarcasm in online conversations - The Conversation

well that’s just great

Here's When You Shouldn't Use AI to Replace Human Judgment - Inc.

Schools turn to handwritten exams as AI cheating surges - Fox News

Companies are doing more personality and cognitive tests in the age of AI. - Business Insider

I’m sure they will be super accurate. You want tests to accurately reflect workplace tasks, maybe they should also do tests of how you use AI to solve problems. Anyway if blue books were good enough for me they are good enough for you young whippersnappers.

Taiwan unveils AI tool to diagnose schizophrenia from brain MRIs with 91% accuracy - Interesting Engineering

AI trap sorts mosquitoes by species and sex for disease tracking - crowsneststpete.com

AI sorter achieves 98% accuracy in carton classification - sustainabilitymag.com

Trust & Will self-help software says its AI upgrade makes it easier to create and update an estate plan. - San Diego Union-Tribune

OpenAI Says It's Hired a Forensic Psychiatrist as Its Users Keep Sliding Into Mental Health Crises - Futurism

Joe Rogan pod tackles AI extinction risk - Gizmodo

The tyranny of Big Tech and of the market. - Honest Broker

George Orwell: In my book I invented the Versificator as a cautionary tale

Gen AI company: At long last, we have created the Versificator from classic sci-fi novel Nineteen Eighty-Four!

Hat tip: https://twitter.com/AlexBlechman/status/1457842724128833538

(Simon Willison)

Hoax band says another hoaxer posing as their rep has nothing to do with their hoax. - CBC

The failure of the AI regulation moratorium appears to have turned on one of the dingbat freak yokels resolve to protect country musicians’ IP, and dedication to child protection regulations. - Slate

"It's not like the president is a baby brained moron," a White House official said, laughing nervously. "He's not, you know, a big stupid asshole who doesn't give a shit about anything except attention" - Alexandra Petri

smooth brain just rattling around in there like Jerry just whacked Tom with a frying pan

Follow the latest AI headlines via SkynetAndChill.com on Bluesky