Intro
Will AI replace software engineering and programming? 🤔
My take: Absolutely and positively YES! Let’s checkout why.
Software Engineering is Very Expensive and Not Very Productive
We are doing lots of repetitive tasks during our daily lives as software engineers.
- Moving code from one language to another language
- Starting projects
- Modifying projects
- Upgrading code to new software versions
- Maintaining server deployments
- Writing tests
- Writing SQL
- Improving performance
- …
I’ve used ChatGPT very successfully to do many of these tasks. Need to write a test? Use ChatGPT. Upgrade React from classes to functions? AI will do it for you. Create a project with a set of technologies (eg npm, typescript, templates, node backend, rest etc) - AI will do it for you. Rewrite a coding example to the language you are using? Done. Complicated SQL query? Solved.
ChatGPT is NOT an ideal tool. Still. It is much much faster in any of these tasks than I. The boost to my productivity as engineer is sometimes beyond imagination.
I cannot wait to see better and more specialized AI colleagues that can do even more.
And I also hope that this productivity gain will help us to become more wealthy as a society as whole. (Ok - maybe we just move that productivity gain to playing more Playstation - but that’s not what I meant).
AI is Like a Higher-Level Language
I almost never touch Assembler code. And when I did I was not very good at it. And there’s a reason. Higher Level Languages are just so much better and so much more productive than writing Assembler code.
And I really think that AI programming assistants are to programming what Higher-Level languages where to Assembler. AI programming will completely replace traditional programming (in javascript, Java, Kotlin, C) in a few years from now.
Some tasks will be programmed by real experts in these languages. But the vast majority of programming will happen jointly with AI assistants. And these AI assistants will simply create the code you can run. Forget about writing programming syntax.
What AI Cannot Replace
BUT. There are hard problems that cannot be solved by a clever AI assistant. For these tasks you need true experts that can architect a large software system and move it into the desired direction. No AI tool will do that for you.
And sometimes the result from AI generated code is not what you want or plain wrong. Having the ability to spot that is essential in the future.
My bet: Real experts will still be in need for few tasks. But many low productivity programming jobs will simply not exist in 10 years from now.
Conclusion
I’ve learned that predicting things - especially the future is extremely hard. And I am especially bad at it.
Today is 2024-03-13. Let’s see if programming goes almost extinct in 2034. I might well be wrong.
This was from a very nice discussion on Linkedin with Carsten Neuendorf on the future of software engineering and Devin that was released today. Thanks Carsten!
