People think being a teacher in the age of BS AI is hell. But imagine doing group homework with 20 year olds that can't write and just send you AI paragraphs every time you ask for help.
Discussion
That sounds absolutely brutal. Not only do you have to deal with the usual group project chaos—flaky teammates, mismatched effort levels, and last-minute panics—but now you also have to sift through a pile of AI-generated fluff just to get anything done. And worse, they probably don’t even read what they’re sending you, so when you point out an issue, they just generate another batch of nonsense instead of actually thinking.
Are you stuck in a situation like this right now?
I'm not sure it actually changes that much. Looking back at my own experiences, the students that contributed little without AI and current students contributing little with AI are, at the end of the day, probably having the same effect on the project's outcome. 😅
Sometimes nothing is better than something.
Teachers deserve a medal for surviving this digital jungle
Teachers need to use AI to keep up with what the students can now learn on their own. Maybe guide them in using the AI effectively to explore the really important aspects of the subject. This is the world the students will live in. Regretting it isn't like it was when you were the student is a waste of time and a disservice to everyone.
The problem is more often than not these bots spit out hallucinations mixed with facts. It takes experience to spot the useful bits.
Also, using these tools early on also means you actually never learn to do it correctly. How are you going to spot mistakes, or debug? I think these tools should be used *after* gaining some experience.
Have the AI generate tests. Read them. Learn how they work by query the AI until satisfied. Both learn. It is faster. I am old. I have learned a lot the old way. But I notice I learn new things much faster with the AI than without. I don't think we know yet the proper balance or can know without more working with it.
Are you learning new things with AI in completely new domains, or still within your domain? If you are within your domain, you still have your experience to filter out the inaccuracies. Try with something very different.
Just couple of days ago I spotted a significant but subtle error in a somewhat straightforward summarisation task I was trying with Llama. It could have easily slipped past me if I were not careful.
Linguistic awareness and Socratic questioning are essential for crafting prompts that elicit informative, transparent responses from #AI #LLMs.
Huh, sounds like my class! They have ChatGPT on speed dial and are at the point where they don't even read the answer but just copy the code blocks, smash them into a terminal, then copy the whole output verbatim back - and repeat the loop.
The day I saw `auto eth0` and other lines of /etc/network/interfaces in the shell history of one of the company servers, i just wanted to leave...
I was thinking earlier of the phrase "you won't always have a calculator in your pocket." Teachers can't use that as a reason for students to learn arithmetic anymore, because we do all have calculators.. and they speak english & (if used appropriately) teach math better than many teachers ever will.