Offline translation is pretty great. Some image editing tools are pretty great. Games may utilise them in the future. Offline image recognition for searching for images (e.g. "show me pics of grandma"), etc.
It's not particularly widely used now, but the same was true for hardware video encode/decode, hardware accelerated encryption/decryption, etc.
image processing is pretty intense and would likely be handled by the GPU. Efficient embedded NN accelerators like this are meant to be used for more passive things, like noise cancelation or like you mentioned, translation.
I don't know the architecture of AI accelerator in Ryzen processors but I do know a fair amount of image deblurring and denoising tools run on the neural engine on Apple Silicon. The neural engine is good enough for a lot of tasks, provided that your model only uses relatively simple operators and doesn't need full precision.