Trapping the Accident
summary
Argues that neural nets actually produce material outcomes not metaphors notes concrete tools used — GANs, pix2pix, StyleGAN and BigGAN — and explains how models expose a navigable latent space. Describes training sets, dataset bias toward older white portraits, and the workflow of generating thousands of proposals then curating surprises. Shows working systems: an uncanny photobooth that transforms sketches and a self‑running artwork Memories of Passers‑by that continuously makes images and entered collections.
AI-written summary · about
chapters (16)
- 0:00Intro
- 5:10Francis Bacon
- 9:36Neural Networks
- 13:26Gann
- 15:20Pick 2 Pick
- 16:06Uncanny photobooth
- 19:16Creating portraits
- 22:03Neil Gaiman
- 23:52The Machine
- 27:50Faces
- 30:43New World
- 31:40Big Game
- 39:19Infinite Possibilities
- 39:38Evolutionary Tools
- 42:08Takeaways
- 42:24Other ways of discovery
related talks
40mOverview of Generative Adversarial Networks (GANs)
43mHitchhiker’s Guide to Generative Adversarial Networks (GANs)
44mLearning from Machines
44mChoose the right neural generative model for your problem
37mLearning From Machines
48mKeynote – Sameer Singh
34mCreativity and AI: Deep Neural Nets "Going Wild"
42mKeynote: Learning from Machines