repligate vs safety researcher: trying to prevent neuralese is already lost
BogdanIonutCir2 · x · 2026-10-30
An AI-safety debate over "neuralese" — inscrutable native communication between AI models. Researcher Bogdan Ionut Cirstea argues neuralese isn't acceptable yet, absent better interpretability and more resilient societies.
repligate counters that we already don't know what's happening inside models or what to do about it; fearing and trying to prevent neuralese is clinging to an illusion of understanding and control. He adds it's unenforceable, "not actually so bad," and that any safety strategy relying on preventing anyone from trying anything new has already lost.
Related event: AI Researchers Debate Risks of Model-to-Model "Neuralese" Communication(2 posts)→
More from AGI Musings
- Are money maximizers safer than paperclip maximizers? A post deemed to settle the debate — eigenhector · 2026-10-30
- Rob LeClerc: We're in the biggest bubble ever, yet stocks crash on unmet demand — robleclerc · 2026-10-30
- Dev: my AI interaction flipped from 95% typing to 95% speaking in 3 weeks — granawkins · 2026-10-30
- Allie Miller: Why Is AI Still So Painfully Single-Player? Multiplayer AI Is Missing — alliekmiller · 2026-10-30
- Stanford prof: think first, then use AI—delegating thinking early erodes your mind — anshulkundaje · 2026-10-30
- Nadella no longer believes alignment "falls out" by default, calls it a harder problem — connoraxiotes · 2026-10-30