This fragmentation means that parts of computing will progress at different rates. This will be fine for applications that move in the 'fast lane,' where improvements continue to be rapid, but bad for applications that no longer get to benefit from field-leaders pushing computing forward, and are thus consigned to a 'slow lane' of computing improvements. This transition may also slow the overall pace of computer improvement, jeopardizing this important source of economic prosperity.
This is perceptive, but also red ocean thinking. Moving ML to GPU is only bad for other workloads if CPU languishes as a result. But I think there's reason to believe it's a blue ocean. The more performance ML gets out of GPU, the more money the ML business can pour into semiconductors, batteries, EDA, and so forth. CPU benefits from all of these.
I think you can see this at work in the laptop market. Cell phone R&D has driven battery technology. Server R&D has driven efficient high-performance x86 cores. Both have driven power-efficient process nodes. Put them together and you can make incredible laptops.
This is perceptive, but also red ocean thinking. Moving ML to GPU is only bad for other workloads if CPU languishes as a result. But I think there's reason to believe it's a blue ocean. The more performance ML gets out of GPU, the more money the ML business can pour into semiconductors, batteries, EDA, and so forth. CPU benefits from all of these.
I think you can see this at work in the laptop market. Cell phone R&D has driven battery technology. Server R&D has driven efficient high-performance x86 cores. Both have driven power-efficient process nodes. Put them together and you can make incredible laptops.