Is there any scuttlebutt on AMD investing in making their GPUs/ROCm work as seamlessly with the standard ML tooling as Nvidia? It seems like it would be "easie

Is there any scuttlebutt on AMD investing in making their GPUs/ROCm work as seamlessly with the standard ML tooling as Nvidia? It seems like it would be "easier" now that the path has been cleared and all they have to do is reach parity.