Erlang and Deep Learning
summary
Argues Erlang excels as an orchestration and monitoring layer for deep learning by demonstrating a command-line plus web tool that wrapped TensorFlow runs, logged GPU statistics and experiment metadata, and compared runs side‑by‑side. Shows a live 10,000‑step training where accuracy climbs, cites lineage from AlexNet to ResNet (50–152 layers), and highlights surprising compute costs (hundreds of GPUs, thousands of CPU cores). Presents the open Guild AI project as a working Erlang prototype for long-running orchestration, collected GPU stats, 10,000 training steps, guild AI project.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:38Deep Revolution
- 1:15AlphaGo
- 2:13Selfdriving cars
- 3:08reinforcement learning
- 3:55machine learning
- 5:05neural network
- 7:22training mechanism
- 11:10the payoff
- 15:15nonlinear functions
- 20:30the revolution
- 23:35the cost
- 25:48Nvidia
- 28:26assembler optimizations
- 29:00tensorflow
- 32:18fault tolerant
- 32:49Python dominates
- 35:10Distributed Deep Learning
- 36:07Gilday
- 37:14Demo
- 42:00Questions
- 43:52The Perceptron
- 44:37John Costas
related talks
24mKeynote: Next-generation frameworks for Large-scale AI
43mDeep learning beyond the learning
49mDistributed Deep Learning
25mApplied Machine Learning in Elixir
23mOperationalizing Large Language Model Training Pipelines
23mBest Practices for Building Production-Grade Deep Learning Systems
36mSkyRL: Democratizing RL Training for Researchers
53mMachine Learning Platforms