AI Reading for Monday April 21
Watch humanoid robots run alongside humans, with a few falls along the way - YouTube
Only half finished the course, and they still walk like they shit their pants, despite all the fluent demos which I’m sure are not cherry-picked. I think 10 years for robots with full bipedal mode and finger dexterity. But more better robots on wheels with vision and AI and on-screen human avatars you can talk to will probably gain more adoption soon.
China is living a few years in the US’s AI and robot future, for better or worse - the Guardian
The Intelligence Community's AI Revolution. The CIA would never abuse AI though? - Reason.com
Robo-dozers offer autonomous home-destroying capabilities - The Economic Times
Can dispatch them with ChatIDF, or automatically with an agentic workflow.
‘Don’t ask what AI can do for us, ask what it is doing to us’: are ChatGPT and co harming human intelligence? - the Guardian
Instagram is testing AI to flag teen users and enforce safety settings. - Fast Company
but mah engagement?
OpenAI teases software agent A-SWE - ZDNET
As it pursues Windsurf, which calls itself an agentic IDE with agents assisting on your software, running in the background and suggesting fixes.
The race to turn brainwaves into fluent speech - FT
To Make Language Models Work Better, Researchers Sidestep Language - Quanta Magazine
make them reason in 'neuralese' before generating tokens
Harrison Chase (LangChain) wrote a long blog post on agents with a spreadsheet comparing them - LangChain Blog
Multi-Tool Orchestration with RAG approach using OpenAI's Responses API
When you use AI to respond to a scam-bot.
Follow the latest AI headlines via SkynetAndChill.com on Bluesky