Deep Learning as a Scientific tool

Deep Learning has been popularized by large companies such as Google, Facebook, Amazon, Microsoft, etc. In doing so, they oriented the field towards the most meaningful applications for them:

• Sound processing (Speech-to-text for mobile phones)
• Computer Vision (Content moderation, Face recognition, etc.)
• Natural Language processing (Automatic Translation, content classification.)
• Recommender Systems (Search results, twitter/facebook feed reranking, content recommendation, etc.)

Today, most resources (teaching, code, business applications examples, etc.) are centered on these applications, occulting some of the newer.

This is however just a small part of what the Deep Learning is impacting, and can be impacting.

• Content generation (Image generation / rendering ; text generation ; voice generation)
• Deep RL (for robotics control, game playing)
• Scientific advances (protein folding, fluid simulation, quantum wave function simulations, etc.)

In this post, I’ll focus on the last one and try to demonstrate why it is meaningful and promising.