Low code, no code and IA generated code

by Laurent Cetinsoy published the 08/08/2022

Low code, no code and machine learning generated code

If you are an anxious developer, you might feel that this time it is your turn. What do I mean ? There is a pretty hot topic lately in “tech” called “low code” or “no code”. These seducing terms try to convey the idea that, now yes, we promise ! This time tools are good enough to let you build applications without coding. Or maybe with just a tiny amount of code, at most ! You the sale person, you the VP,  you can build your own sales application tailored for your own needs. You, the data analyst, you can build your own vertical specific data analysis application without paying a costly developer  (and maybe creating alongside a gigantic mess of  poorly maintained applications in the IT system, but still ) ! 

The anxious developer could hear instead “well you invested a lot of years into your craft but, sorry, you are expensive, go click on some nice interface to develop my fancy application 10 times quicker”.

Would it happen like that though ? Theses tools are always somewhat limited and developers are far from being extinct !

The anxious developer can, nevertheless, grow his fear once more by considering the latest advances in machine learning based natural language processing, like GPT 3.

Usually natural language processing is used on, well, natural language like French or English. Its goals is to automatically analyze and extract meaning from text documents. For example, it is used for hateful comment detection on social networks, to grade student on national exams (not a too fancy idea) and many other things.

But these techniques can be adapted for artificial languages. And I do not mean elfic ! The most used man-made crafted languages are programming languages and have nice names like python, java or Cobol. Such techniques applied to programing language have great potential for bug catching or code generation and thus, lower code !

The most basic use case is to have an AI based code generator embed in your favorite code editor, say vim, which would provide an “intelligent” code completion based on your own code. You can also use machine learning to generate things that developers hate : unit tests. 

Who hates regular expressions ? Everyone ! Why not use ML so it writes it for you ? People might also like the idea of getting rid of SQL : you can now ask an IA a question in plain english and it generate the SQL query. There are a lot of examples made with a famous method called GPT3

So coding will be made easier in a lot of fields and this is a good thing ? Does-it mean that developer will be gone forever ? Unlikely for now, like many other ML based systems, it will enhanced worker productivity and human in the loop is still needed. The sometime anxious developer I am is relieved. However, bear in mind that theses systems are in their infancy and should better over time.

 

One additional thought : In my opinion, non tech people are not that much interested to develop their own applications, low code or not, they want their problems solved.