~www_lesswrong_com | Bookmarks (705)

Published on May 22, 2025 1:40 PM GMTWhat a week, huh? America signed a truly gigantic chip sales agreement with UAE and KSA that could be anything from reasonable to civilizational suicide depending on security arrangements and implementation details, Google announced all the things, OpenAI dropped Codex and also bought Jony Ive’s device company for $6.5 billion, Vance talked about reading AI 2027 (surprise,...

Published on May 22, 2025 1:18 PM GMTBPC-157, a peptide frequently marketed as a breakthrough for healing and tissue repair, has attracted substantial attention in wellness and performance communities. It’s discussed in forums, recommended by biohackers, and even offered in clinics—despite lacking FDA approval or broader clinical recognition.The challenge is personal: we all want healing when we’re hurting—but how do we evaluate bold health...

Published on May 22, 2025 12:08 PM GMTMotivation One major risk from powerful optimizers is that they can find "unexpected" solutions to the objective function, which score very well on the objective function but are not what the human designer intended. The canonical example is Suppose your aged mother is trapped in a burning building, and it so happens that you're in a wheelchair;...

Published on May 22, 2025 11:10 AM GMTCore Argument The barrier to preying on "mirror-image" life is likely lower than presumed due to: Immediate energy from achiral fats. Existing L-sugar pathways. Existing D-amino acid racemases. 1. Fats: Achiral Energy Bridge Lipids (triglycerides), largely achiral, offer immediate energy. In E. coli for example, they constitute ~10% of dry mass, yielding ~27% of its macronutrient calories[1])....

Published on May 20, 2025 6:30 PM GMTWhile we wait for wisdom, OpenAI releases a research preview of a new software engineering agent called Codex, because they previously released a lightweight open-source coding agent in terminal called Codex CLI and if OpenAI uses non-confusing product names it violates the nonprofit charter. The promise, also reflected in a number of rival coding agents, is to...

Published on May 20, 2025 6:09 PM GMTWe will soon enter an unstable state where the balance of military and political power will shift significantly because of advanced AI.As different nations gain access to new advanced technologies, nations that have a relative lead will amass huge power over those that are left behind. The difference in technological capabilities between nations might end up becoming...

Published on May 20, 2025 4:21 PM GMTWelcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required.In this edition: The Trump Administration rescinds the Biden-era AI diffusion rule and sells AI chips to the UAE and Saudi Arabia; Federal lawmakers propose legislation on AI whistleblowers, location verification for AI chips,...

Published on May 20, 2025 3:34 PM GMT2025-05-20 WARNING This guide is currently a research draft. DO NOT follow the guide as of today, if you are a whistleblower. I do not yet endorse following this guide as being sufficient to protect your safety. I am only sharing this to get research feedback. I will update you once I do think the guide is...

Published on May 20, 2025 3:26 PM GMTIn the comments section of You can, in fact, bamboozle an unaligned AI into sparing your life, both supporters and critics of the idea seemed to agree on two assumptions:Surviving planetary civilizations have some hope of rescuing planetary civilizations killed by misaligned AI, but they disagree on the best method of rescuing.The big worry is that there...

Published on May 20, 2025 2:13 PM GMTOn May 20, during her speech at the Annual EU Budget Conference 2025, Ursula von der Leyen, President of the European Commission, stated:When the current budget was negotiated, we thought AI would only approach human reasoning around 2050. Now we expect this to happen already next year. It is simply impossible to determine today where innovation will...

Published on May 20, 2025 12:54 PM GMTWe study how selective regularization during training can guide neural networks to develop predictable, interpretable latent spaces with alignment applications in mind. Using color as a test domain, we observe that anchoring even a single concept (red) influences the organization of other concepts, with related concepts clustering nearby — even with weak supervision. We then propose that...

Published on May 20, 2025 9:07 AM GMTComputers get smarter. People don't. Some bots will be greedy and some will not. The greedy ones will take everything. Discuss

Published on May 20, 2025 6:40 AM GMTYou can read Ten arguments that AI is an existential risk by Nathan Young and I at the AI Impacts Blog.Discuss

Published on May 20, 2025 6:40 AM GMTHave the Accelerationists won? Last November Kevin Roose announced that those in favor of going fast on AI had now won against those favoring caution, with the reinstatement of Sam Altman at OpenAI. Let’s ignore whether Kevin’s was a good description of the world, and deal with a more basic question: if it were so—i.e. if Team...

Published on May 18, 2025 4:34 PM GMTWith Canada's new Government and Cabinet in place, we must act quickly and decisively in the global AI race. To ensure Canada leads and that AI development benefits all Canadians, urgently securing resources and fostering partnerships is critical. I've sent the following letter to our new Minister of Artificial Intelligence and Digital Innovation today, hoping to spark...

Published on May 18, 2025 4:22 PM GMTLook, nothing I can say about GEB will be news to anyone reading LessWrong, but I wrote all this up to organize my thoughts, so I might as well post it here for anyone who hasn't read the book or wants a wave of nostalgia about that time they got obsessed with it as a teenager. I...

Published on May 16, 2025 4:55 PM GMTThis is a crosspost from the Southern China Morning Post Few expected it would happen this fast, but there is widespread relief at the news that China and the United States are talking again. In only a few days, renewed trade talks have resulted in temporary but large reductions of the mutual tariffs that seemed so impregnable just...

Published on May 16, 2025 4:36 PM GMTPhenomenology is a philosophical discipline that seeks to understand and describe human experiences as they are perceived. It always was a difficult topic to explore as people struggle to even understand what is being discussed; it does not attempt to study conscious experience the way psychology or neurology do. Instead, it seeks to determine the essential properties...

Published on May 16, 2025 4:10 PM GMTThe system prompt being modified by an unauthorized person in pursuit of a ham-fisted political point very important to Elon Musk once already doesn’t seem like coincidence. It happening twice looks rather worse than that. A Funny Thing Happened on Twitter In addition to having seemingly banned all communication with Pliny, Grok seems to have briefly been...

Published on May 16, 2025 11:59 AM GMTThis post is a reflection on criticism raised by a brief critique of reduction. I've thought it would be too big for a comment so here it goes... I've read all of the posts from Reductionism 101 and Joy in the Merely Real and enjoyed my time. But I think that a brief critique of reduction was misunderstood...

Published on May 16, 2025 5:09 AM GMTLanguage models are not particularly good at generating funny jokes. Asked for their funniest jokes, Claude 3.7 gives us:Why don't scientists trust atoms? Because they make up everything!o3 gives us:Why don't scientists trust atoms anymore? Because they make up everything—and they just can't keep their quarks straight!and Gemini 2.5 Pro gives us…Why don't scientists trust atoms? Because...

Published on May 16, 2025 3:44 AM GMTStatus: Very early days of the research. No major proof of concept yet. The aim of this research update is to solicit feedback and encourage others to experiment with my code. GitHub repositoryMain Idea:Researchers are using SAE latents to steer model behaviors, yet human-designed selection algorithms are unlikely to reach any sort of optimum for steering tasks such...

Published on May 16, 2025 2:20 AM GMTNote: This post is the second in a broader series of posts about the difficult tradeoffs inherent in public access to powerful open source models (the first post is here). While this post highlights some dangers of open models and discusses the possibility of global regulation, I am not, in general, against open source AI, or supportive of...

Published on May 15, 2025 10:29 PM GMTWithout going into the philosophical or technical underpinnings too much, several individuals I know from a wide range of backgrounds—computational neuroscience, computer science, philosophy, AI research—have put together a project where we hope to put some real energy towards studying consciousness in synthetic environments. Obviously, if this is an area of interest, you'll likely have questions about whats...