I think it depends.

On one side, I do agree that having a system that works with typos and incomplete text ends up with the user writing incomplete inputs filled with typos...

But on the other side, I've found that the best types of answers you get from these LLMs, is when you actually write a well crafted text, in which you specify all the important things that matter to you. I guess this might be what people call prompt engineering. When you're interacting with the LLMs in this way, it actually makes you think a lot more about what you're asking, and makes you write a much better question, which sometimes makes you get to your answer even before sending the prompt as you figured out while thinking it through.