Robot on Rails
summary
A presentation-generator bot is trained on 14,000 SlideShare presentations (about 100 megabytes of slide text) using an lstm recurrent neural net / torch RNN running on GPU instances (EC2, jutsu 2x large); training takes roughly 24 hours with checkpoints every few minutes, early output is gibberish, after ~30 minutes it learns word and slide tropes and later produces occasional coherent nuggets, so the demo pipeline is generate → sample checkpoints → human-select fragments and paste into slides, and it requires a GPU or rental EC2 instance to be practical.
AI-written summary · about
chapters (4)
related talks
01h 29mBuild Your Own Data Science AI Agents
01h 01mKeynote: Transformers: The Rise of ChatGPT
16mKeynote: Robot Foundation Models
53mGenerative AI: Overview and use cases
49mKeynote 4: Here Come The Robots - Python and Machine Learning
29mBuilding Machine Learning demos with Python
57mWhat if Picasso was a robot? Teaching art to the machine
28mDelivering AI on Code: Live Demo of source{d}