AI, the end or new beginnings?

AI has made it easy for quick fixes into code. As a maintainer of several project repositories, it feels like a relief. Lots of frameworks to assist in coding. I know what I want to do and AI feels like a little helper to quickly type certain code segment that I understand. It also helps find mistakes in code quickly. Similar to what IDEs did, but with extra support. But as an educator, I also feel dangerously horrible use of AI. Students are just throwing essay titles into ChatGPT. When an assignment slightly deviates from the norm and when they cannot let AI to complete it for them, some students are feeling frustrated. Students (both local and those who participate in remote programs - so the issue is global. Not limited to Alaska) simply let AI do coding and writing for them, and when something goes wrong, they are confused. For a trivial fix that would take the code author just a couple of minutes and a couple of lines to fix, the AI ends up with drastic changes. 

Sunset in Barbados

Many student essays feel bland. They lack soul and personality. If you give the student an well-designed homework, they quickly get the answer. When the problem involves thinking on architecture beyond what AI can handle for them at this point, they are stuck. "Can you tell me what I should exactly do here?" For vaguely defined problems, they want clear step-by-step instructions which they can then easily feed into LLMs.

As educators, we should spend more time in teaching architecture and thinking. If all students want is to simply feed clear instructions into ChatGPT, if I were an employer, I would rather do that myself. Why would I pay someone when they need clear instructions, step-by-step guidelines to a software program that they can easily feed into ChatGPT or other more advanced AI coding tools?

I used to enjoy fixing student papers. Those essays had life and personalities. They read like what students wrote. They sounded like the student. Now, they all sound bland. Like an AI slop. And they are indeed detected at 50 - 100% as AI by AI detectors of tools such as Grammarly Premium. At this point, I could simply write everything on my own. That will be, starting from zero, as opposed to trying to fix AI slop, which feels like starting from -100.

Computer science undergraduates are giving up the fight to LLM even before they start. This is sad as this is also a self-fulfilling prophecy.