Computing inside an AI

Where simulation is good enough, it is transforming robotics. But it’s nearly impossible to engineer simulators that can accurately model complex physical and visual properties.

How you can get a 100x speedup for training small neural networks by making the most of your accelerator.

Cross-posted from the CILVR blog. In the last few years, there's been an explosion of work on learning good representations of data. From NLP123 to computer vision456 to reinforcement learning789, the field has never been hotter. However, defining precisely what we mean by a good representation can be tricky. This has led to a somewhat ad-hoc...