Software 2.0, a Babel fish for deep learning?
summary
Claims that mathematical models written as Unicode formulas can be parsed and transpiled into an execution graph that separates the learned weights from the runtime engine. Shows a working proof of concept: a Go-based parser that reads math notation and emits a graph, so models become data plus graph rather than hand-coded layers. Demonstrates parsing examples, notes limitations, and reports a runnable prototype rather than just an abstract idea.
AI-written summary · about
chapters (4)
- 0:00Start
- 0:54Different programming languages for different ways to describe an algorithm.
- 2:35To write machine learning model, you need to understand: • Mathematics AND • Programming language oddity • Cloud computing • Web services Infrastructure
- 4:07For more information and a POC, cf the Gorgonia Project (coded in Go)
related talks
31mThe Evolution Revolution
48mMachine learning by an imp-op-steur
05mTalks with Brian Troutwine about His Code Mesh Talk
37mNeural Networks are Programs, Too
49mOpen Models: The Secret Weapon for Next-Generation Software
36mIt's Alive! Machine Learning Writes Your Code! 🧠🔮
01h 10mModels and Programs: Better Together
29mLearn Neural Networks With Go, Not Math