PinnedJason Robinson·Apr 15NetHack Is Still Unbeaten. That Should Worry You.NetHack came out in 1987. It runs in a terminal, renders in ASCII, and fits on a floppy disk. No AI has ever beaten it.
PinnedJason Robinson·Apr 24Play Is How Systems May Learn What Can’t Be FoundA child stacking blocks is generating facts through interaction with objects whose behavior was never cataloged. Each stack that stands and…
Jason Robinson·Apr 30Where to Focus in the AI AgeIf you write software for a living, one part of what you do is being automated quickly and the other part is not. The part that is being…
Jason Robinson·Apr 21Proofs Compound. Pattern Matching Doesn’t.Scaling improves pattern matching, and the AI industry has organized around scale as the path to capability. What scale produces and what…
Jason Robinson·Apr 17AGI Has Been Here Since 2020AGI arrived in 2020. It was called GPT-3. The few who called it AGI at the time got argued with until they stopped saying it out loud.
Jason Robinson·Apr 16Proof-Based AI: What It Is and Why It Is ComingIn January 2024, DeepMind published AlphaGeometry in Nature. It combined a language model with a symbolic deduction engine and solved 25…
Jason Robinson·Apr 14Safety Is Architecture, Not TrainingA customer support team deploys an LLM-based agent. It works. Three months in, the vendor updates the underlying model. Refusal rates…
Jason Robinson·Feb 2The Knowledge Machine: What If AI Could Learn Without Us?Every major AI system trains the same way. Collect data. Label it. Have humans rank outputs. Train a reward model from those rankings…
Jason Robinson·Jan 29Structural Alignment: Why Training AI to Be Good Isn’t EnoughDario Amodei’s recent essay “The Adolescence of Technology” is the most serious treatment of AI risk from an industry leader to date. He…A response icon1A response icon1
Jason Robinson·Dec 17, 2021The value of Python is the velocity of idea to implementation.A response icon1A response icon1