How do you believe prominent deep learning ASIC makers such as Nvidia and Google will alter their hardware design over the next decade to meet the exponentially expanding need for extra computing to train and execute machine learning models?
For example, since the beginning of the current machine learning boom, the industry has adopted numerous of architecture changes such as the incorporation of mixed-precision Tensor based processing units into their hardware. What other changes that is reminiscent of this change can the industry implement?
