Andrew Ng on Deep Machine Learning via Large Scale Brain Simulation
https://www.youtube.com/watch?v=5elcmFNRCWk
The key determinants of learning accuracy are accessing as much data as possible and being able to process.
Partnered with the Google speech team.
Deep learning works well in two different settings:
1. Learning from labeled data > speech recognition, streetview images
2. Learning from unlabeled data >
Up to now, humans have been driving performance due mainly to a lack of data and processing. With both of these now becoming available in abundance, machine learning will soon overtake human learning to become the dominant driver of performance.
Sky net.