There are still a lot of software side of things we as a community can do in accelerating compute speed - for example, add OpenCL support for popular AI framewo

There are still a lot of software side of things we as a community can do in accelerating compute speed - for example, add OpenCL support for popular AI framework/libraries, which will leverage existing hardware and non-Nivida GPUs' power. [1]: https://github.com/tensorflow/tensorflow/issues/22 https://github.com/tensorflow/tensorflow/issues/22