Will software that writes code alter tech’s script?

“We learnt that we were building the same thing over and over, and realised that the more interesting thing to do was to cut out the wasted effort and cost, and work out how to automate 90 per cent of writing code.”

I view that as a failure of abstraction and composition, not as a failure of automation. If it's tedious to hook up window controls to underlying data, or to handle properly a form validation loop in a controller for the millionth time, we should be asking why the abstraction layer we are using requires that tedium instead of being more succinct?

Often the abstraction is complex because of the degrees of freedom it has. Automation on the other hand reduces the degrees of freedom. This is great if you are the one who is reducing and making the choices, but if you are letting someone else make all the choices about what is possible... It's like the difference between driving and being chauffeured.

Even if this system were like the mathematician: "show me all possible worlds satisfying X" it would require either a huge search space to pick the desired answer or present only a handful of preselected choices. Search refinement is more interesting, sounds like a great deal of complexity to avoid (less?) complexity of actually doing the work-- especially in meanial cases-- i.e. I don't want to sift through 1000 variations of form validation to "find" the magic one that completely meets my business needs.

Here's a good comment from a reader, which I think addresses a misunderstanding behind this article about what programming actually is:

> The thing that takes time is humans understanding what they want. Programmers are like therapists who help them crystalize their thoughts to the point where they are at least vaguely consistent. The program is just a representation of that idea.

We already have code that writes code. Guess who uses it: programmers.

It doesn't matter how smart the computer gets, a smart programmer with the computer is always going to be more useful than the computer on its own.

I think having the computer translate your natural language intentions into code is at least as hard as translating your natural language intentions into a different natural language. I don't see that problem being solved to sufficient accuracy, so I really doubt that this approach will replace traditional programming in the next ten years.

The shark example brings to mind Star Trek TNG, when someone would 'write' a programme by simply giving vague voice commands to the computer, and end up with exactly what they wanted.

In reality it's very difficult to describe visual things in words.

Also, how does a fierce shark look different to a shark?

Ha. Having conquered the paywall:

Ridiculous. COBOL. AppleScript. Lego Mindstorms. Also:

Is it difficult to develop a programming language which is closely related to human language? (http://programmers.stackexchange.com/questions/267489/is-it-...)

Or, in one of the clearest and most succinct pieces Dijkstra ever wrote:

On the foolishness of "natural language programming".

https://www.cs.utexas.edu/users/EWD/transcriptions/EWD06xx/E...

I doubt if anyone could describe in natural language a complex enough idea clearly and consistently without doing it.

It's not AI, it's Super Intelligent.

So first they try and get me to download an app and when I decline, I get a paywall. I suspect the content isn't with the cost or effort to work around it.

This day finally comes... yet.

lol... this is Andrew Ng being asked whether machine learning will solve software anytime soon:

http://www.reactiongifs.com/r/tmimn.gif