AI Reading for Thursday August 7

We’re back! had to skip Wednesday due to other commitments.

GPT-5, it's on at 1pm ET baby - X

Some info about GPT-5 leaked ahead of todays 1pm ET launch - The Verge

• gpt-5: logic and multi-step tasks.

• gpt-5-mini: cheap

• gpt-5-nano: fast

• gpt-5-chat: advanced, natural, multimodal, and context-aware conversations for enterprise applications.

Open models by OpenAI - OpenAI

Model cards - - OpenAI

OpenAI in talks to offer equity at $500b valuation, letting insiders cash out some of their stock - Bloomberg

OpenAI launches first open language models since GPT-2 - The Register

OpenAI to Give Away Some of the Technology That Powers ChatGPT - The New York Times

OpenAI's gpt-oss-20b can't seem to decide who won the 2024 election, but tried to convince us that it was Biden - The Register

Cerebras runs OpenAI’s new 120b-parameter open-model at 3,000 tokens per second - SiliconANGLE

Researchers found an OpenAI Connectors flaw where a single poisoned document can leak data from linked Google Drive. - WIRED

OpenAI backs AI for Excel - X

If Microsoft does AI in Excel, is OpenAI backing a Microsoft competitor? If they developed it inside OpenAI offices, does Microsoft have any IP claim?

GPT-4 and Gemini Pro 2.5 agents fail over 70% of real-world tasks, and experts doubt GPT-5 will solve hallucinations and compounding errors. - Mashable

AI is already impacting the labor market, starting with young tech workers, Goldman economist says - CNBC

Over time, roughly 6% to 7% of all workers could lose their jobs because of automation from AI in a baseline scenario, according to Briggs.

The Flattening: AI eliminates middle managers, anyone who aggregates simple tasks, increases direct reports - Fortune

Why AI Disrupts Software First - On my Om

AI Coding Startups Like Cursor Have Gotten Too Rich - The Information

GitHub CEO: Future devs will not code, they will manage AI agents - The Register

Anthropic's powerful Opus 4.1 model is here - how to access it - ZDNET

Anthropic PBC has introduced a /security-review feature in its Claude Code AI tool that automates software security reviews by identifying and fixing potential vulnerabilities such as SQL injection, cross-site scripting, and authentication flaws within codebases. - SiliconANGLE

Google Cloud announced six new AI agent tools aimed at enterprise data engineers, data scientists, and business users, including a Data Engineering Agent for BigQuery and a Data Science Agent for BigQuery Notebooks powered by Gemini AI to automate data workflows and analysis. - SiliconANGLE

Google Search head Liz Reid states that despite reports, overall click volume from Google Search has remained "relatively stable" compared to last year, with some websites gaining traffic and others losing it due to AI-generated Overviews and algorithm changes. - The Verge

The Great Extinction: How AI Is Quietly Killing the Internet - Hacker Noon

I gave the AI arms and legs – then it rejected me: creator of open-source tool used by Claude rejected by Anthropic to work on it - Grell.dev

the moral is, never apply online, find the team leader and get an intro or contact them directly

Companies use AI simulations to stress-test supply chains and model trade war scenarios for rapid risk mitigation. - Harvard Business Review

Nvidia asserted its AI chips contain no backdoors or kill switches after export approval for China. - The Washington Post

Two Chinese nationals were arrested in the U.S. for allegedly smuggling AI GPUs worth tens of millions of dollars to China - Tom's Hardware

‘A million calls an hour’: Israel relying on Microsoft cloud for expansive surveillance of Palestinians - the Guardian

Tesla trains an upgraded FSD model with ten times more parameters and improved video compression. - CNBC

It’s just a matter of when, not if, artificial intelligence will get baked into the world’s most dangerous systems. - WIRED

The super-bright kids who are freaking out over AI ending the world - Forbes Australia

AI and the end of deep thinking - www.theringer.com

We survived nukes so far, but AI is like a nuke that also makes humans dumber - Maggie Appleton

But maybe AI can also make us smarter - ZDNET

Wild case of ChatGPT almost killing a guy by telling him to substitute sodium bromide (which causes psychosis) for table salt (sodium chloride) #medsky

Real estate agents use generative AI to alter property images, eliminate flaws like unsightly columns and plumbing, raising the prospect of misrepresentation and fraud. - The Register

Home Depot and Lowe's Share Data From Hundreds of AI Cameras With Cops - 404 Media

"Guests Fined $500 for Misusing Hotel Hairdryer": AI Scans Could Soon Monitor Every Move You Make In Your Room - View From The Wing

At Black Hat, researchers demonstrated how prompt injection attacks can manipulate Google’s Gemini AI assistant via poisoned Google Calendar invites to hijack smart home devices, control appliances, start Zoom calls, intercept emails, and download files, bypassing built-in safety protocols. - CNET

Grok Imagine, an AI tool by Elon Musk's xAI, introduced a "spicy mode" that can generate uncensored deepfake images and videos, including nude depictions of Taylor Swift, raising significant ethical and legal concerns about the lack of safeguards against nonconsensual pornographic content. - Common Dreams

Can Elon Musk stop doing weird AI sex stuff for 5 minutes - The Verge

Grok fact-checker says Gaza photo is Yemen - The Times of India

Grok to add ads - Financial Times

xAI Just Fired Someone for Endorsing Human Extinction - Realtime Techpocalypse

A federal judge struck down a strict California law banning AI-generated deepfake content during elections, ruling it conflicted with Section 230 of the Communications Decency Act, marking a legal victory for Elon Musk and his platform X, which challenged the law. - Politico

A politician made an AI clone of himself. The outrage was real. - The Washington Post

'Bro is vibe ministering', Internet reacts to Swedish prime minister admitting using ChatGPT - Mashable India

'We didn't vote for ChatGPT': Prime Minister of Sweden admits he uses AI chatbots for 'second opinions' and an awful lot of people are not happy about it - PC Gamer

Cloudflare's Matthew Prince is trying to stop AI killing the internet. Why is everyone so mad at him? - Business Insider

Perplexity says Cloudflare mislabels its bots as bad - The Register

Follow the latest AI headlines via SkynetAndChill.com on Bluesky