Machines that write
Trained on a trillion words of human thought, language models complete essays, contracts and confessions — one probability at a time.
[ Chapter 03 — Capabilities ]
Six disciplines, six generative portraits — each artwork on this page is drawn live by the very kind of algorithms it describes. Move slowly. Tilt the cards.
Trained on a trillion words of human thought, language models complete essays, contracts and confessions — one probability at a time.
Convolutional eyes find edges in chaos: tumors in scans, pedestrians in fog, your face in a crowd of eight billion.
AlphaFold folded two hundred million proteins — a century of biology compressed into a weekend of computation.
Voice becomes text, text becomes symphony. Diffusion models hallucinate frequencies no orchestra has ever played.
Copilots autocomplete software the way you finish sentences. Roughly half of all new code is now co-authored with a machine.
From warehouse arms to humanoid prototypes, reinforcement learning turns simulation into muscle memory for machines that touch the world.
“The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.” — Edsger Dijkstra