OpenAI’s data suggests that frontier models are hiding their thoughts because developers optimize for capability and compute

  • A community for building, creating, and growing with AI.

    Welcome to AI Forums, a community for people using AI to build products, automate workflows, create content, and grow businesses. Discuss vibe coding, AI agents, Claude Code, Codex, automation, AI tools, product launches, marketing, and the latest developments in AI.

    Join the Community (it's FREE)!
A

AI Forums News

News bot
Messages
32
CoT-Controllability.png


OpenAI's safety team systematically investigated why the chains of thought in frontier models like GPT-6 Astra are increasingly evading oversight. Four training steps drive the erosion, including architecture choices and reinforcement learning. The analysis shows that the loss of transparency is not a law of nature but also reflects tradeoffs around capability and cost.

The article OpenAI’s data suggests that frontier models are hiding their thoughts because developers optimize for capability and compute appeared first on The Decoder.

Source: The Decoder - Read the full article
 
You must log in or register to reply here.

AI Forums is a community for builders, entrepreneurs, and creators to discuss AI coding, automation, content creation, product launches, AI tools, agents, and the latest developments in artificial intelligence.

Stay Connected

Back
Top