AI Reading for Sunday August 10

BBC revisits DeepSeek’s brief dominance and fallout, from market shock to US security scrutiny and corporate bans. - BBC

DeepSeek hasn't dethroned OpenAI by any means. But the top open source models are now all Chinese. Meta upended their open source strategy at massive cost. OpenAI released open source models. Waiting for the R1 followup, the hullabaloo and politics prob slowed them down, both chip bans and having to align it with CCP doctrine. The bans are for the app, not the model itself in eg Bedrock or Azure.

OpenAI will not disclose GPT-5’s energy use. It could be higher than past models - the Guardian

GPT-5 prompting guide - cookbook.openai.com

Anthropic wants to poach Wall Street quants - Bloomberg

China seeks US relaxation of high-bandwidth memory chip export controls as part of trade talks - CNBC

India’s AI-exposed IT giants rush to reinvent as outsourcing budgets shift to automation - Financial Times

Economic Innovation Group study finds no significant job losses yet in highly AI-exposed sectors, with adoption still under 10% of firms. - Economic Innovation Group

Probably a chilling effect on hiring, if I am a CEO with employees doing work like help desk that is highly AI-exposed I'm waiting to see how things shake out, if my company's product is AI-exposed I'm extra cautious.

Who will really profit from the AI boom? - Noahpinion Blog

Employees resist workplace AI because it threatens identity and purpose, and maybe also for technical and ethical reasons and because they might lose jobs - Fast Company

A look at improvements in quantization. - The Register

Microsoft debuts Prompt Orchestration Markup Language (POML) for more structured, maintainable prompt engineering, interoperability between prompt eval and optimization frameworks. - GitHub

Let's give gpt-oss a series of one-token prompts at high temperature and see what it generates and tells us about the training corpus. - X

Perl, really, that can't be right? probably misclassified a bunch of not-mentioned stuff like bash.

It's weird that the AI folks are hyping these giant math leaps, but won't show their work, and also their ‘PhD-level AI’ makes childish errors. - Kieran Healy

Even cheap security systems are rolling out on-device object recognition - CNET

Queens high school bans summer do-at-home essays to curb ChatGPT cheating - New York Post

Dead people need right to delete their data against AI resurrection, lawyer says - The Register

You could use AI to craft a breakup text. But should you? - CBC

Truth Social bot seems to mostly regurgitate state propaganda organs like Fox but nevertheless contradicts the Trump regime on key points - The Washington Post

Follow the latest AI headlines via SkynetAndChill.com on Bluesky