Trinity of Artificial Intelligence
summary
Argues that modern AI needs three concrete ingredients: ImageNet 14 million images and 1,000+ categories for data, deep learning models that reach human-level accuracy on ImageNet by 2015, and GPU parallelism >1 billion ops per image for compute. Shows tensors used to decompose word triplet co-occurrences and power a deployed system called Amazon Comprehend (topic extraction from millions of documents). Demonstrates learned drone landing succeeds where the preprogrammed controller only hovers, and proposes fusing instinct, planning and multi-agent behavior for future robots.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:48What is Intelligence
- 1:35Robot vs Dog
- 2:33Intelligence
- 3:25General Intelligence
- 4:13Trinity of AI
- 5:01Image Classification
- 5:59Image Net
- 6:49Deep Learning
- 7:59Computer Infrastructure
- 10:59Learning in Many Dimensions
- 12:28Tensors
- 12:51Text Document Categorization
- 13:41Amazon comprehend
- 14:39Mind and body
- 15:29Learning
- 16:14Drones
related talks
13mArtificial General Intelligence: Why aren't we there yet?
18mOn the History and Future of AI
59mThe evolution in AI thinking and products of the next decade
48mArtificial Intelligence. Our journey towards singularity
40mMachines are Learning
29mDeep Learning, Neuroscience and the future of AI
45mArtificial intelligence: How data is evolving the future of technology
04mAI and the Robotics Revolution