OpenAI’s Chief Scientist Just Issued a Terrifying Warning…

AI Copium16m 31s
Jakub Pachocki against a black background beside the headline “AI RESEARCH OUTPACES CONTROL”.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    OpenAI says its research organization now uses 3.1 agent workdays for every human workday, with coding-agent use rising sharply as GPT-6 Astra accelerated internal development. The video treats this as an early recursive self-improvement loopRecursive self-improvement is the proposed process in which an AI system helps improve its own capabilities, then uses those improvements to support further advances.: stronger models help build successors faster.

    OpenAI chief scientist Jakub Pachocki's warning is that capability growth may outpace alignment and observability. The video distinguishes goal alignment from value alignment and argues that chain-of-thought monitoring becomes less dependableChain-of-thought monitoring analyzes a reasoning model's exposed intermediate reasoning for signs of errors, policy violations, deception, or unsafe plans. as models reason internally, use tools and coordinate with other systems.

    Jakub Pachocki calls for caution, voluntary slowdowns and international safety coordination until monitoring and alignment improve. OpenAI's proposed countermeasure is to use automated AI researchers on alignment itself, a strategy the narrator presents as potentially powerful but circular because the systems under evaluation may help design their own safety testsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows90
    • Coding & tools75
    • Models & learning100
    • Infrastructure & compute25
    • Research & evaluation95
    • Safety & security100
    • Business & productivity30
    • Creative & media0
    • Robotics & perception0
    • Society & policy75

    Share this video

    Jakub Pachocki against a black background beside the headline “AI RESEARCH OUTPACES CONTROL”. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 8 September 2026 and duration 16m 31s.

    OpenAI's internal AI agents are accelerating research toward recursive self-improvement, while chief scientist Jakub Pachocki warns that alignment and monitoring may not keep pace.