Quality

Why should we learn anything? That is a waste of time. Every two-hours-task is done in 5 minutes now. Why could we even be bothered? Why should we even have classes and teach people any kind of proper crafting?

Well, there is quite a number of multiplex reasons.

Proper maintenance is only possible by understanding.

The topic, as we could expect, is not new. Every time there is a new technology, we try to automate the work of people. It already starts with hammers, screwdrivers and assemblers. And of course, we don’t want to write machine code and turn in screws with bare hands – but, the thing is – we should be able to.

Why?

What happens, if we buy any kind of machinery and use it productively, although we don’t understand how it works? If it doesn’t work as expected, we cannot adapt it, let alone fix any problem by our own. Like if my car stops working in the middle of nowhere, I can’t fix it without a mechanic. The same counts for software that nobody in your company has built.

What does the inventive but highly priced person do better?

It is possible that the generative approach leads to the same outcome, although unlikely. But if some kind of ideation produced it, there is a model behind it that is a kind of understanding that goes way beyond just documentation. Even more so, if the model is produced by collaborative work and kept alive over its cycles and surviving fluctuation. But there is more. It doesn’t even matter if the generated code is only aggregation of information found in the web and old practice or new idea by non-humans, supposed that the later will be even possible with old information as a basis.

If any business lets complexity be managed by black boxes, it won’t be in control of itself or any of its products.

Real testing is learning

GenAI also produces tests and metrics guarantee code coverage automatically. What could possibly go wrong?

Where to start? How do we define coverage and is it to be trusted? Of course there are good aspects about covering test sets, but that is not sufficient. Most of the defects in software are the product of leaving something out. Your bug is most likely something that is missing. No coverage can find missing pieces for you, by definition tests cover existing parts of the software.

Also, the existence of automated tests is not the only purpose they need to fulfill. Tests that are written during the creation of a work piece are means of understanding the problem, no generation can replace that.

The generation’s generation problem

Yes, it gets worse. Looks like the trend shows that not only we are replacing experienced people with low-cost solutions, but we are also replacing inexperienced people with automation as they are not standing out from the results produced by automation. Causing not only a competition problem that hinders collaboration, but also – and more importantly – leading to a system when nobody learns building stuff by themselves. This causes educational systems to degenerate, making every problem even longer lasting and more difficult to solve in the future.

So, now, more than ever, it is more important to sit together and live though a proper process of building something than just have the outcome ready to use.

Reference: https://www.spf-consulting.com/en/insights/teaching-the-mob-to-students/