I get his point, but this really seems like another case where writing the ChatGPT prompt takes almost as much effort as writing the message yourself.
What if programmers of the future have almost no knowledge at all of how a program works? I guess in much the same way as most modern programmers have no knowledge of machine code, but even more so. They don't even know what language they're writing in.
"Write an app that connects people of similar interests, and incentivizes them to sell products to each other."
"Make it friendlier-looking"
"Add Stripe payments"
Anyone can do it, but just like artists of the future will be those who have memorized all the clever phrases to get Stable Diffusion to do its thing, programmers will be wizards who know all the invocations that generally seem to make something that works roughly right, and doesn't have too many bugs.
(Edit: Asking ChatGPT to "write JavaScript program that asks the user for their name and then greets them with a funny nickname it invents based on the name" resulted in a perfectly decent program.)
It's not far from it now. I have it spit out components and I don't know all the details. It just works
I don't think so. Crafting good requirements that guide the AI to the solution will be a dominant skill. If you can't articulate your problem and possible solutions, the model isn't going to be able to do anything for you.
I am senior engineer that works with skilled customers and they have a difficult time even understanding the problem let alone the solution to the problem they think they have.
Chatting AIs up all day long will improve people's vocabulary and expression. Language models could be the best language learning partner.
Functionally illiterate here means that there is a reduced reliance on reading, writing or on the end user's ability to understand the code produced.