p(doom)
pdoom.org · 195 words · saved by 1 readers
Turning sand into intelligence. One bitter lesson at a time.
Turning sand into intelligence. Addressing the fundamental bottlenecks toward AGI that scaling compute cannot solve. Unblocking the exponential. Meta-Learning from Humans Unlock data troves that existing training paradigms cannot reach. These trajectories capture people as they learn: acting, receiving feedback, adapting, and improving over time. Month-Long Horizons Learn, reason, and act across trajectories that last hours, days, and eventually lifetimes. Fixed-Size State Carry useful state indefinitely while keeping memory and computation bounded as experience accumulates.…
saved by
related reading
- Physical Intelligence (π)pi.website
- The Future Worth Building Is Human - Thinking Machines Labthinkingmachines.ai
- Varieties Of Doomminihf.com
- Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Labthinkingmachines.ai
- Doomprompting Is the New Doomscrolling.workingtheorys.com
- I. From GPT-4 to AGI: Counting the OOMs - SITUATIONAL AWARENESSsituational-awareness.ai
- LLM Powered Autonomous Agents | Lil'Loglilianweng.github.io
- Varieties Of Doom — LessWronglesswrong.com
- Breaking the Intelligence Curse - The Intelligence Curseintelligence-curse.ai
- A Pragmatic Vision for Interpretability — AI Alignment Forumalignmentforum.org
- What we’d like to fund — Paradigm 3paradigm3.org
- ACT-1: A Robot Foundation Model Trained on Zero Robot Data | Sunday Robotics | The helpful robotics companysunday.ai