Steal this talk
summary
Built a pipeline called Talkie that downloads YouTube captions, stitches Elixir and Python binaries to train a recurrent neural network on hundreds of Ruby talks, and generates titles, descriptions and full talk text; loss drops toward ~1 on small datasets and produced plausible titles like “The Impermanence of Ruby.” Also creates RSS podcast feeds of audio at 3× speed and a cloned-voice avatar — generated talks are coherent enough to submit, and the voice clone works but feels creepy.
AI-written summary · about
chapters (6)
- 0:00Intro
- 7:57Recurrent Neural Networks
- 12:45Neural Networks
- 14:28Backpropagation
- 15:25Perceptron
- 22:55Tips
related talks
39m"Steal This Talk"
43mGiving my wife a voice with Ruby and AI cloning (+ discussion)
35mThe Podcast Pipeline Nobody Sees: Behind the Scenes with AI Agents and AI gen
21mChalkTalk: Globalizing Data Science Education with AI-generated Videos
29mHello, computer. Writing Ruby with voice recognition
16mToo Much News, Not Enough Time? Generate Your Own Watchlist Podcasts!
20mKnowledge Scout: From Talks and Slides to Smart Summaries
27mI know Kung Fu! RubyGems for AI
