AI Reading for Friday May 16?

Hmmh, pasting images no longer works? Hello? Sorry, no time for this. Been meaning to possibly move to Ghost.

---

Trump’s Middle East tour is all about AI diplomacy - Fast Company

---

House GOP Move to Block State AI Laws Sounds Cyber Alarms - WSJ

---

China hawks think that if the US sells $10b of AI hardware to Gulf states, it will get laundered to China - Axios

---

China Accuses US of ‘Unilateral Bullying’ on Huawei AI Chips - Bloomberg

---

Internal Microsoft memo reveals plans for a new ‘Tenant Copilot’ and an ‘Agent Factory’ concept - Business Insider

---

OpenAI adds GPT-4.1 to ChatGPT amid complaints over confusing model lineup - Ars Technica

---

ChatGPT is testing MCP integration, presumably similar to how the Claude desktop client can connect to any MCP server and use it as a tool. - Bleeping Computer

---

Claude AI is getting a big Reasoning upgrade: Everything you should know - BGR

---

Llama 4 seen to have fallen off the pace as largest model is now delayed - Business Insider

---

Meta to postpone release of Llama 4 Behemoth model, report claims - SiliconANGLE

---

Google DeepMind creates super-advanced AI that can invent new algorithms - Ars Technica

---

Why We’re Unlikely to Get Artificial General Intelligence Anytime Soon … There are some hardcore scaling law true believers and what we have already is immensely impactful, but most people think we will need some more breakthroughs beyond LLMs - NY Times

Big debate between doomers who say AI will kill us all, deniers who say AI is all hype (deny what they see with their own eyes), and dreamers who say we will have 10% GDP growth and cure all the diseases and do 100 years of science in 10 years.

---

HN discussion of AlphaEvolve and its improvement to Strassen’s matrix multiplication algorithm - HN

Hard problem a lot of people looked at

I’m not sure how practically important it is, there is asymptotic complexity and average complexity, numerical stability etc

this sort of incremental improvement is one thing but to do 100 years of science in 10 years, you need to find radically new stuff

breakthroughs like relativity are a search problem but on a very high conceptual level that I don’t know if LLMs can do.

they might ideate and ask good questions though

---

Databricks to buy serverless cloud Postgres startup Neon for $1 billion to boost AI-agent development - CNA

---

Chain gang: LangGraph Platform for cloud deployment reaches general availability - VentureBeat

---

Ollama adds multimodal local models now that they are supported by underlying llama.cpp - Ollama

---

Vibe-coding startup Windsurf launches in-house AI models - TechCrunch

---

Samaya AI, startup selling AI tools to Wall Street, gets $43.5 million in new funding - Fortune

---

An interactive AI tool reveals how companies respond to economic threats. Make your own Google Trends over e.g. earnings calls. - TechXplore

---

AI chatbots compete over memory and personalization - FT

I’ve gotten rejections, “I can’t do that”, started a new chat, and noticed that o3 traces interpreted the new question by referring to the previous rejection]

---

YouTube is turning to AI to hit you with ads at the most annoying times - BetaNews

---

Is Opera Leading the AI Agent Browsing Era? - IEEE Spectrum

---

If AI is so good at coding … where are the open source contributions? - Pivot to AI

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky