This is a very hardware focused overview of technology. To summarise, Moore's Law is dead. Specialise processors takes over. And they will be owned by specific Companies.
That is all true as I have been posting pretty much exactly the same on HN for a few years. The real question is though, does it matter?
I am not convinced, at least not yet convinced what we are doing now on a computer is that much different to 20 years ago. Excel, Web Browsing, Media Consumption and Content Creation like Photoshop. The only real recent innovation is Machine Learning that greatly increase productivity in certain niche.
And it does not seems to me any of the above activity is going to fundamentally shift. There is nothing in terms of Hardware tech limitation that is holding up performance for further productivity increase, rather it is Software that is not getting much improvement if you consider the worst and best software could have performance different of 100x.
To give an example, I could bet with few billion dollar we could create a Hardware GPU that is 80% close to Nvidia GPU or AMD GPU. I am also willing to bet even with 10 Billion dollar funding we cant get a CUDA Compatible Library or Drivers that is 80% of what Nvidia is offering today.
That is all true as I have been posting pretty much exactly the same on HN for a few years. The real question is though, does it matter?
I am not convinced, at least not yet convinced what we are doing now on a computer is that much different to 20 years ago. Excel, Web Browsing, Media Consumption and Content Creation like Photoshop. The only real recent innovation is Machine Learning that greatly increase productivity in certain niche.
And it does not seems to me any of the above activity is going to fundamentally shift. There is nothing in terms of Hardware tech limitation that is holding up performance for further productivity increase, rather it is Software that is not getting much improvement if you consider the worst and best software could have performance different of 100x.
To give an example, I could bet with few billion dollar we could create a Hardware GPU that is 80% close to Nvidia GPU or AMD GPU. I am also willing to bet even with 10 Billion dollar funding we cant get a CUDA Compatible Library or Drivers that is 80% of what Nvidia is offering today.