This sounds like the account of someone that hasn’t actually used the technology recently

I’ve not had any misunderstandings by Google assistant or Alexa in ages, and LLM produced code works great for me now provided you’re explicit enough in the way you instruct it (meaning you still need to understand what you want it to do)

Image generation is great for if you want a quick illustration, inspiration or a meme, not much practical benefit besides that imo