The Post-It Note on the Vault | The Truth About AI Guardrails
Send us Fan Mail
In this episode of Mind Cast, host Will dives deep into a groundbreaking research paper titled "Geometric Encoding of Asimovian Invariants in Open-Weight Neural Architectures." We unpack why current post-training safety techniques—like RLHF, DPO, and fine-tuning—are fundamentally fragile behavioural filters easily bypassed by hobbyists using consumer hardware. We then explore a radically new approach to AI alignment: mathematically encoding Isaac Asimov’s Three Laws of Robotics directly into the high-dimensional geometry of neural networks as load-bearing structural…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.