How a Data-Centric approach drives reliability

Andrew Ng:
Deep learning is not only expanding the set of applications that machine vision can now address, it is also improving the reliability of these systems. And the key to that is the Data-Centric approach, in which getting the right images to show to the system is what allows it to derive those reliable and repeatable results. Further, one of the long-standing goals of a data-centric strategy is to reduce the time to deploy and time to scale these systems, and really the core part of the secret sauce of the deep learning approach is that rather than needed to program the systems step by step, all you need to do is show the system images, and with the right tools they’ll come up with the right images to show the system. That then drives the reliability, repeatability, as well as a fast and easy development, deployment, and scaling of the process.