If we can get GPUs deployed at the scale we want, adapting to whatever language or framework you want to use to program them will be easy. The challenge is gett

If we can get GPUs deployed at the scale we want, adapting to whatever language or framework you want to use to program them will be easy. The challenge is getting GPUs into enough machines in enough places to make this interesting. But, again, we're working on it.