It is not easy to do "prompt engineering". Keep in mind that most generated material looks awful -- there is a huge selection bias going on.
The outputted stuff needs to made production ready also, distribution, etc.
If anything, the emergence of these technologies should greatly expand. You will have a new customer base that was not available before, and you will be able to deliver new products.
Yeah. Look at machine translation - it's gotten pretty good, in some limited cases it's perfect, but it's also frequently just clueless, and for anything serious you absolutely need a real human translator.
I think these AI tools will improve, but I'm extremely skeptical that it's going to improve by like an order of magnitude any time soon. It will still require extensive tuning, heavy selection, and accepting that you probably can almost never get it to make exactly what you want.
Another example is the uncanny valley of CGI people: at what point was it acceptable to audiences, and at what point was it excellent?
It's pretty early to declare that a victory for human drivers isn't it? I'd be worried, if I were them.
You spelled "two years" wrong.