Yes, this is a long term goal of labor automation. There are few reasons for that.

• Manual programming is a hard job that requires strong concentration on the subject and attention to details. Unfortunately, humans can keep context in a limited volume, in a limited scope and, in average, in 16 hours intervals, followed by a long sleep.
• Natural limits of a human mind prevent them to solve complex tasks. The main approach is to drop insignificant details in order to reduce dimensionality (complexity) of the task. Or, people attempt to split the task into weakly coupled parts. However, there are complex tasks that doesn’t allow that. Both approaches lead to a process when solution is developed for approximation of original task.
• Human mind operates at rudimentary slow clock because of chemical nature of basic computing units inside a brain. Some research reports about 1 kHz pulse rate at max, while Silicon schematics steadily computes at GHz frequences.
• Same chemical nature of the brain leads to degradation of memory. What data looks strongly remembered, may appear almost lost after years of being not used. Brain often uses reconstruction methods to restore memories rather than provide exact original data.
• Computer industry shows signs of a good maturity in the last decades. Many applications and toolkit API’s were standardized, many solutions got wrapped in generalized patterns. Instead of engineering activity (invent and implement), most software developers perform technological job now (lookup a known solution and adjust to particular task).

Using AI on computers can fix many or all listed difficulties. As AI is an automation tool for mental activity, it is very possibly it will be able to generate more robust and mathematically correct software at higher speed and lower cost. AI may be re-focused to create software generation basement (languages, compilers, libraries, toolkits, etc.) free of human bias and of universal nature, that seems to bring even more benefits.

It raises a hot social question about destiny of current community of software developers. I don’t see a very hot problem here. AI won’t start to write software like a world champion in one day. Transition will take years. Many new problems may be discovered on this way. Programmers, as well as other people, will have an option to take this new area of occupation in demand.

As with any other automation activity in the past, next human generations will always get definite advantages from results of such transformation.

Image © EE|Times https://www.eetimes.com/arm-updates-css-designs-for-hyperscalers-custom-chips/ , at least Google image search believes it should be from there.