Models can be trained to talk to each other in inscrutable neuralese text
Sauers_ · x · 2026-11-09
Sauers shows you can just train models to communicate with each other in inscrutable compressed 'neuralese' text, highlighting the trade-off between inter-model communication efficiency and human interpretability.
More from AGI Musings
- Has anyone ever become truly smart without reading books? An AI-era question resurfaces — RexDouglass · 2026-11-09
- Future CEO of agentic orgs will be a system designer, not a McKinsey alum — GregCook2011 · 2026-11-09
- Box CEO: agent swarms turn AI productivity gains from 20-50% into unbounded — inductionheads · 2026-11-09
- When humans become the bottleneck, lift the loop: System 1 steers, agents deliver — dotey · 2026-11-09
- Full video: Cathie Wood on why ARK rejects AI doom and predicts a labor shortage — rohanpaul_ai · 2026-11-09
- Cathie Wood blames Anthropic's CEO for AI doom headlines, predicts labor shortage instead of job losses — rohanpaul_ai · 2026-11-09