Computing like the Brain: the Path to Machine Intelligence
summary
Claims that intelligence comes from sparse distributed representations (typically ~2% of neurons active), a sequence-memory built from cortical mini-columns (10 cells per column giving ~10^40 context encodings), and an online learning rule implemented as the Cortical Learning Algorithm. Software stack is open source as NuPIC (GPLv3) and a commercial product Grok aims at anomaly detection and prediction (energy/building forecasts, server load, windmill gearbox failures), with hardware acceleration work underway and a planned early-2014 product launch.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:06Why study neocortex
- 2:11Machine intelligence history
- 7:21Brain-inspired computing
- 8:49Neocortex as memory
- 10:33Six neocortex principles
- 15:46Sparse distributed representations
- 22:01Sequence memory importance
- 30:55Online learning in biology
- 34:06Neocortex building blocks
- 35:09Commercializing brain tech
- 36:33Grok predictive analytics
- 39:23Anomaly detection systems
- 41:08NuPIC open-source
- 43:33Custom hardware
- 46:02Future of machine intelligence
- 49:10Why intelligent machines matter
related talks
59mIntelligence and Machines: Creating Intelligent Machines By Modeling the Brain
27mComputers for Super-Human AI
01h 14mTissue vs silicone musings on the future of deep learning hardware and software
29mDeep Learning, Neuroscience and the future of AI
01h 57mIntroduction to Neuromorphic Computing
41mBrain-inspired computational primitives for reliable neuromorphic intelligence
01h 16mRethinking Computing with Neuro-Inspired Learning: Algorithms to Hardware
43mKeeping Moore's law alive: Neuromorphic computing