No, because we already have a machine API. If you want to write an application, you need to write something a computer can understand. So a computer-usable API is always created. It takes additional effort to hide that functionality behind a interface. The process we have now is: machine → GUI → image processing → generative AI. The interface we could have is: machine → machine. It would take no extra effort to do this. It would just need some slight changes in organisation. In fact it is easier at every level. If you separate logic from interface, you end up with an architecture that is a set of functions (a library) into which you can interface programmatically, or with a GUI, or by any other means. Separating code like this (MVC) is good practice and allows for a range of different interfaces to be created to the same functionality. It is also easier for an engineering perspective and produces a better product. Think of git. There are hundreds of different interfaces created to the functionality git provides. All software should be structured like this (though perhaps by means of a library rather than a shell interface).

I should add that this is a particularly grim prospect from a software engineering perspective. It makes me imagine a future where no one bothers exposing a stable API to anything, so the only way to interact with other people's code is using an AI middle-man.