Conclusions
Deep Learning is enabling a new wave of applications

• Today: Image recognition, video understanding: vision now works
• Today: Better speech recognition: speech recognition now works
• Soon: Better language understanding, dialog, and translation
  

Deep Learning and Convolutional Nets are being widely deployed

• Today: image understanding at Facebook, Google, Twitter, Microsoft…..
• Soon: better auto-pilots for cars, medical image analysis, robot perception
  

We need hardware (and software) for embedded applications

• For smart cameras, mobile devices, cars, robots, toys….
  

But we are still far from building truly intelligent machines

• We need to integrate reasoning with deep learning
• We need a good architecture for “episodic” (short-term) memory.
• We need to find good principles for unsupervised learning