Jackson Kozlowski

on the atlas — 91

• The Hugging Face attack surprised me - by Ajeya Cotra3 savers
• Why I Left Anthropic - by mrinank - living more whole1 savers
• What is input/output filtering in AI safety? - by Sarah2 savers
• Introduction to AI Control - by Sarah - BlueDot Impact2 savers
• Using Dangerous AI, But Safely? - YouTube2 savers
• Buck Shlegeris on controlling AI that wants to take over – so we can use it anyway | 80,000 Hours1 savers
• When AI Chooses Harm Over Failure - CivAI2 savers
• AI models can be dangerous before public deployment - METR2 savers
• We Need A ‘Science of Evals’1 savers
• AI 2040: Plan A1 savers
• Please don't throw your mind away — LessWrong30 savers
• A simple technical explanation of RLH(AI)F | Kairos.fm2 savers
• Problems with Reinforcement Learning from Human Feedback (RLHF) for AI safety2 savers
• A small number of samples can poison LLMs of any size \ Anthropic6 savers
• Enhancing Model Safety through Pretraining Data Filtering2 savers
• Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs1 savers
• What is input data filtration in AI safety? - by Sarah1 savers
• How far does alignment midtraining generalize?3 savers
• What is AI alignment? - by Adam Jones - BlueDot Impact3 savers
• China’s AI Regulations and How They Get Made | Carnegie Endowment for International Peace1 savers
• AI Act Single Information Platform | AI Act Service Desk1 savers
• Dario Amodei — We Must Pace the Frontier31 savers
• Anticipatory governance | OECD1 savers
• When Reporting an AI Security Incident Is Not Mandatory | Lawfare1 savers
• The Best Charity Isn't What You Think2 savers
• Expected value: how can we make a difference when we're uncertain what’s true? | 80,000 Hours1 savers
• Scope Insensitivity - LessWrong4 savers
• Common Elements of Frontier AI Safety Policies - METR1 savers
• Why I Don't Buy Any of the Counterexamples to Consequentialism1 savers
• Notes on not taking the GWWC pledge (yet) — EA Forum1 savers
• Dedicated Donors May Not Want to Sign the Giving What We Can Pledge — EA Forum1 savers
• Contra the Giving What We Can pledge — EA Forum1 savers
• Nobody Is Perfect, Everything Is Commensurable | Slate Star Codex10 savers
• AI Safety's Biggest Talent Gap Isn't Researchers. It's Generalists. — EA Forum3 savers
• Trees are mostly made of air and a generalizable lesson for AI safety  — LessWrong10 savers
• Useful Vices for Wicked Problems7 savers
• Concrete Generalist Projects in AI Safety (and how to do them) — LessWrong1 savers
• How to be right about ethics when it matters - by Flo Bacus1 savers
• Increasing Our 2026 Allocation to GiveWell’s Recommendations to $1 Billion | Coefficient Giving1 savers
• An Ethical Diet1 savers
• The case for AI safety capacity-building work — EA Forum2 savers
• Rest Days vs Recovery Days - LessWrong4 savers
• AlgoTransparency1 savers
• Being the (Pareto) Best in the World - LessWrong11 savers
• Writing, Briefly2 savers
• Good Writing5 savers
• Having Kids17 savers
• How to Lose Time and Money7 savers
• OpenAI Email Archives (from Musk v. Altman and OpenAI blog) — LessWrong2 savers
• Elon Musk and OpenAI - Internal Tech Emails1 savers
• stanford masters program - Google Search6 savers
• Why Charities Usually Don't Differ Astronomically in Expected Cost-Effectiveness — EA Forum1 savers
• Commitment ability in multipolar AI scenarios – Center on Long-Term Risk1 savers
• A high-level model of AI bargaining — LessWrong2 savers
• Our position on open-weights models \ Anthropic7 savers
• Cooperation, Conflict, and Transformative Artificial Intelligence: A Research Agenda – Center on Long-Term Risk1 savers
• CLR's Safe Pareto Improvements Research Agenda — LessWrong1 savers
• Mini Blog Post 3: Become a person who Actually Does Things — Neel Nanda45 savers
• The Soul of EA is in Trouble - by Matt Reardon1 savers
• Scholarship: How to Do It Efficiently — LessWrong2 savers
• The Neglected Virtue of Scholarship — LessWrong2 savers
• Learning By Writing42 savers
• Minimal-trust investigations9 savers
• You Can Learn Any Subject by Trying to Write About It | by Ron Markley | Medium1 savers
• Neural networks and deep learning2 savers
• Neural networks and deep learning11 savers
• Neural networks and deep learning3 savers
• Philosophers Are the Latest Hiring Target for AI Companies - The New York Times1 savers
• One Minute in Hell - Matt Beard's Substack1 savers
• ARENA 7.0 Impact Report — LessWrong1 savers
• On Doubling9 savers
• Architecting Local Incentives - Madeleine1 savers
• How to Make Wealth6 savers
• The third wave of American philanthropy - by Nan Ransohoff17 savers
• Everything I love is downstream of powerful AI1 savers
• Curius / Onboarding2621 savers
• Circuit Tracing: Revealing Computational Graphs in Language Models20 savers
• Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident - METR20 savers
• What will be scarce? - by Alex Imas - Ghosts of Electricity14 savers
• A retrospective of AI alignment14 savers
• Discovery of a new OpenAI agent message board12 savers
• Eliezer's Unteachable Methods of Sanity — LessWrong11 savers
• Detecting misbehavior in frontier reasoning models | OpenAI7 savers
• Summary of METR's predeployment evaluation of GPT-5.6 Sol7 savers
• How I Work - by Dean W. Ball - Hyperdimensional6 savers
• The choices we make about AI now are critical | Bill Gates4 savers
• Opinion | The A.I. Giants Weren’t Prepared for This - The New York Times3 savers
• A gentle introduction to sparse autoencoders — LessWrong3 savers
• Deferral is a skill - by Alex Lawsen - Speculative Decoding2 savers
• Security Level 5 - Nation-State Grade Security for Frontier AI2 savers
• The self-actualiser vs the hedonist2 savers