We really need some competition to Nvidia's CUDA ecosystem. Both Intel and AMD have some half-baked initiatives, but they need to ramp that stuff up quickly (as

We really need some competition to Nvidia's CUDA ecosystem. Both Intel and AMD have some half-baked initiatives, but they need to ramp that stuff up quickly (as quickly as one can catch up to 15 years of development, while the competition was complacently sitting on OpenCL). We can't have all the new AI models be powered by Nvidia, that clearly won't end well. I really hope it won't be nvidia and cloud exclusive ASICs like Google's TPUs.