As +Mark Cummins said, a talk full of wisdom. Note that Andrew Ng says (around 2:15) that deep learning only works because it is able to take advantage of huge amounts of data -- Geoff Hinton has said similar things -- so it's really the data paired with an algorithm that is able to take advantage of the data (deep learning isn't interesting on small data, and other algorithms may eventually catch up like they did in the 1980s). Also don't miss the bit at the very end (around 1:05:00) where he's talking about what deep learning can do well, in particular, tasks where labeling can be done very quickly (< 1 sec) so enough data can be created.
This talk is so full of wisdom. Lots of great high level advice about how to build deep learning systems and run an ML organisation.
Shared publiclyView activity