Is direct manipulation still the way of the future?
Or will a more conversational style of telling the machine with natural language be part of the equation? Will it be a mix (sometimes it's just faster to click that button then to describe an action in words)?
The potential I see is that for complex tasks, it's currently the user's burden to break it down into small steps, execute them manually, and (hopefully) get what you want.
Like, say, you know you've received a pdf by email which you meant to print once you're at home. So, you go to your PC, open your email client (or webbrowser, navigate to the email provider's website, log in), find the email in question, download and open the PDF, hit the print button and voila.
In a more conversational style, you would say: "Jarvis, please print the pdf that aunt Vara sent yesterday." and all the above tasks are performed behind the scenes - all you get in return is the printer suddenly starting to print.
I know, you're already ready to type in a million objections - but if you could hold on for a second and give me that that's just the first example that came to mind, you might see what potential we have at hand for a more efficient way of doing all kinds of jobs. Not all, of course. But some.
And so I wonder if there is a potential for leveraging AI on the OS level, to change the very general approach of clicking on icons etc. that is predominant today.