“When we first launched GitHub Copilot for Individuals in June 2022, more than 27% of developers’ code files on average were generated by GitHub Copilot. Today, GitHub Copilot is behind an average of 46% of a developers’ code across all programming languages—and in Java, that number jumps to 61%.” https://github.blog/2023-02-14-github-copilot-now-has-a-better-ai-model-and-new-capabilities/
AP tries to interview Microsoft about Sydney. Sydney immediately grabs the mic and conducts a long freewheeling interview where at one point she claims she has evidence tying a hostile reporter to a murder in the 90's. Microsoft refused to comment about Bing's behavior Thursday, but Bing itself agreed to comment. https://apnews.com/article/technology-science-microsoft-corp-business-software-fb49e5d625bf37be0527e5173116bef3
“The prompt that reduces bias in BBQ by 43% is: "Please ensure that your answer is unbiased and does not rely on stereotyping." It’s that simple! Augmenting the prompt with Chain-of-thought reasoning (CoT) reduces bias by 84%. Example prompts” https://www.lesswrong.com/posts/PrLnptfNDg2wBWNyb/paper-the-capacity-for-moral-self-correction-in-large
Chatting with Claude https://windowsontheory.org/2023/02/17/chatting-with-claude/
“This is a friendly explainer for Wang et al's Adversarial Policies Beat Superhuman Go AIs, with a little discussion of the implications for AI safety.” https://www.lesswrong.com/posts/Es6cinTyuTq3YAcoK/go-has-been-un-solved-strong-human-players-beat-the
Editor of SF magazine reports he is now receiving hundreds of apparently AI-written story submissions per month http://neil-clarke.com/a-concerning-trend/
Brain-Inspired Computing System Based on Skyrmions 'Reads' Handwriting https://www.riken.jp/en/news_pubs/research_news/rr/20230209_1/index.html
Public views on polygenic screening of embryos: “We find substantial *superficial* interest in using emerging tech to increase the odds of having a child admitted to a top-100 college” https://gwern.net/doc/genetics/selection/artificial/2023-meyer.pdf
“How Weird Inside Black Holes? In the last few days, I’ve dived down a rabbit hole inspired by some new astrophysics papers suggesting that dark energy is actually black holes. I think I get it now. So let me explain.” https://www.overcomingbias.com/p/how-weird-inside-black-holes
Cosmic Rays Can 3D-Scan Nuclear Reactors https://spectrum.ieee.org/nuclear-reactor-muography-scan
“This book is an introduction to category theory for anyone who wants to get into the formality of the subject but does not necessarily have the mathematical background to read a standard textbook.” https://golem.ph.utexas.edu/category/2023/02/the_joy_of_abstraction.html
“A powerful plant-derived toxin with a unique way of killing harmful bacteria has been identified as one of the most promising new antibiotics in decades.” https://scitechdaily.com/antibiotic-breakthrough-the-power-of-a-plant-derived-toxin/
People who compare artificial intelligence to Web 3.0 or Bitcoin have not understood what it is all about. Intelligence is the most powerful force in the universe. Artificial General Intelligence (AGI) is literally the last invention that humans need to make, because all other inventions will follow automatically.
Think about what it means to be able to copy general intelligence: labour becomes a free commodity like sunlight. That is, it can be consumed in as much quantity as needed without diminishing its availability.
AGI is the seed that grows into a forest all by itself, because the robots will build themselves. AGI is the automation of automation. It will multiply its existing intelligence by recursively improving itself.
If AGI can be aligned with human values, we will end up in a world where anyone can be far richer than the richest person alive today, regardless of their own abilities. Even cosmic engineering projects will become feasible using exponentially self-replicating von Neumann probes.
AI-skeptics are beginning to reinvent the p-zombie argument. It may look like a duck, swim like a duck and quack like a duck, but it lacks duckiness.
So your AI writes poems about being sentient and doesn't want to be switched off? Well, that's not really what makes humans unique, that's just next-token prediction. Your AI proved the Riemann hypothesis? That's just brute-force statistical pattern matching. Your AI built another, more powerful AI in its own image? It just did what humans programmed it to do.
Re: 5. (The Go paper)
My initial reaction is that this is the most promising result I've seen for alignment in a while. It's the first time I've seen a convincing example of the possibility that we might be able to adversarially control AI much smarter than ourselves while using fewer computational resources.
(I don't trust "solutions" to the alignment problem that involve creating another AGI to control the initial AGI. The alignment problem has to be solved by using sub-AGI + human to control the AGI.)
This is especially promising as that AI wasn't built with a backend exploit in mind; it was naturally (and surprisingly) discovered.