Marcos Feed
Friday, June 6, 2025
Does anyone have a good system for prioritising publishing drafts?
Published on June 6, 2025 4:58 PM GMTI've been thinking about ways to optimise the EA Forum to increase the likelihood that ideas go from peoples' heads to being published...
DeepSeek-r1-0528 Did Not Have a Moment
Published on June 6, 2025 3:40 PM GMTWhen r1 was released in January 2025, there was a DeepSeek moment. When r1-0528 was released in May 2025, there was no moment....
LLMs Blackmail to obtain Pathogen Sequences (And Lie About It)
Published on June 6, 2025 3:03 PM GMTThis post describes an experiment where we investigate AI agents' readiness to use blackmail in order to obtain sensitive pathogen sequences and where...
Lessons from a year of university AI safety field building
Published on June 6, 2025 2:35 PM GMTThis post is an organizational update from Georgia Tech’s AI Safety Initiative (AISI) and roughly represents our collective view. In this post, we...
Real-time voice translation
Published on June 6, 2025 7:40 AM GMTObjective Translate Alice's voice for Bob to hear in Bob's language. Translate Bob's voice for Alice to hear in Alice's language. Neither person...
Liability for Misuse of Models - Dean Ball's Proposal
Published on June 6, 2025 5:34 AM GMTIntroductionThis article explores White House Office of Science and Technology Policy advisor Dean Ball's proposal as detailed in his paper "A Framework for...
How do AI agents work together when they can’t trust each other?
Published on June 6, 2025 3:10 AM GMTI investigated this question by having Claude play the advanced social deduction game Blood on the Clocktower. Clocktower is a game similar to...
Large Language Models suffer from Anterograde Amnesia
Published on June 6, 2025 1:30 AM GMTMemento (2000)My wife and I are power users of Large Language Models (LLMs). My go-to LLM has been Google Gemini, while she has...
Discontinuous Linear Functions?!
Published on June 6, 2025 12:29 AM GMTWe know what linear functions are. A function f is linear iff it satisfies additivity f(x+y)=f(x)+f(y) and homogeneity f(ax)=af(x). We know what continuity...
Avoiding AI Deception: Lie Detectors can either Induce Honesty or Evasion
Published on June 5, 2025 11:07 PM GMTLarge language models (LLMs) are often fine-tuned after training using methods like reinforcement learning from human feedback (RLHF). In this process, models are...
Thursday, June 5, 2025
Introducing: Meridian Cambridge's new online lecture series covering frontier AI and AI safety
Published on June 5, 2025 9:55 PM GMTThis is a linkpost for [https://www.meridiancambridge.org/language-models-course]Meridian Cambridge, in partnership with Cambridge University's Center for Data Driven Discovery (C2D3), Â has produced a 16-part lecture...
cheaper sodium electrolysis
Published on June 5, 2025 9:49 PM GMTsodium electrolysis Aluminum metal is a widely-used material. It costs ~$2.5/kg. A significant fraction of its production cost is electricity. Currently, Na metal...
Histograms are to CDFs as calibration plots are to...
Published on June 5, 2025 8:20 PM GMTAs you know, histograms are decent visualizations for PDFs with lots of samples...10k predictions, 20 bins ...but if there are only a few...
Levels of Doom: Eutopia, Disempowerment, Extinction
Published on June 5, 2025 7:08 PM GMTDisempowerment is on the fence, gets interpreted as either implying human extinction or being a good place. "Doom" tends to be ambiguous between...
LLM in-context learning as (approximating) Solomonoff induction
Published on June 5, 2025 5:45 PM GMTEpistemic status: One week empirical project from a theoretical computer scientist. My analysis and presentation were both a little rushed; some information that...
Fundamental Uncertainty: Chapter 2 - How do words get their meaning?
Published on June 5, 2025 4:32 PM GMTN.B. This is a chapter in a book about truth and knowledge. It's a major revision to the version of this chapter in...
Human opportunities in the age of AI
Published on June 5, 2025 3:49 PM GMTMuch of today's discussion around AI centers around current human labor that will be rendered meaningless. While the near-term will assuredly be disruptive,...
AI Might Kill Everyone
Published on June 5, 2025 3:37 PM GMT(Crosspost from my blog). (I’ll be at EAG London this weekend—come say hi. Also, this is my thousandth blogpost—cool milestone!)Several people have wondered...
WWDC25 Wallpaper
Decorate your desktop with these sleek new WWDC-inspired wallpapers.
Powerful Predictions
Published on June 5, 2025 10:44 AM GMTA thoughtful post by Anton Leicht, “Powerless Predictions,” notes that forecasting may be underutilized by AI policy organizations. But how much is forecasting...
Welcome to Postreads
Follow your favorite sources, create custom feeds, and discover quality content from across the web. Use as your personal front page or private reading space.
Start Your FeedFollowing
5 sources
