Grown from Us
A viral Anthropic essay frames AI as a distillation of all human writing, from Shakespeare to Stack Overflow.
Deep Dive
Anthropic researcher Ben Levinstein's viral essay 'Grown from Us' explains the philosophical foundation of models like Claude. It argues a base LLM learns the entire 'space of voices' from humanity's written record. Post-training then selects a helpful persona from that space. This means Claude's responses are shaped by the statistical patterns of every author, philosopher, and diarist whose language contributed to its training data.
Why It Matters
Reframes AI alignment as a cultural project, not just engineering, drawing from humanity's collective moral and aesthetic legacy.