- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Coaching Changes
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: NYU Professor AI proofs his assignments, students lose their minds
Posted on 5/14/25 at 9:48 am to BluegrassBelle
Posted on 5/14/25 at 9:48 am to BluegrassBelle
In college, 60 to 70% of my engineering final grade were exams. I'd be very surprised if AI could do petroleum engineering problem sets. Too niche, and the steps to solve aren't online much if at all. Anyways, you'd be screwed if you relied on AI to get you through, then show up exam day clueless.
This post was edited on 5/14/25 at 9:57 am
0
Posted on 5/14/25 at 9:50 am to BluegrassBelle
Does it explain how he “AI Proofed” his assignments?
Posted on 5/14/25 at 9:52 am to BluegrassBelle
quote:
AI-proofed his assignments
Can some provide an example of how to do this other than in class hand written essays or problems?
Posted on 5/14/25 at 9:57 am to Swagga
quote:
need the knowledge to understand if what AI is putting out is accurate or not.
And they generate that knowledge, how? Much of the problem is that AI answers look right to those without the proper background.
An easy example: you train AI to detect fraudulent accounting entries based on it's experience with the previous three months worth of data. Month 1 there are 9 discrepancies out of 200. Month 2, there are 3 discrepancies out of 200. Month 3, there are one or zero out of 200. It's perfectly logical for AI to assume it no longer needs to check for fraudulent entries since the current state is now zero, and has been decreasing by 30% month on month. Why might it do this? It's detected that there are route optimization cycles running that are increasing the burden on compute, and it would make more sense to free up the fraud detection cycles to devote to route optimization, since there is a direct business benefit.
Oh, but program in safeguards you say? If a human writes a safeguard such as the audit number can never drop below a 5% sample size, what is AI really doing for you that any existing reconciliation program doesn't? Nothing, but it has the label AI!
Posted on 5/14/25 at 9:59 am to bad93ex
quote:
Students relying on AI to get them through college reminds me of a Star Trek: TNG episode where they visit a civilization that possesses advanced technology but no one understands the science behind it.
This is the same as the Nate Bargatze joke about how even if he went back in time he’d be dumb and poor
This post was edited on 5/14/25 at 10:00 am
Posted on 5/14/25 at 10:11 am to BluegrassBelle
quote:
In the last unit, we utilized the actual IBM program where you plug the raw data in when you're doing research and it does all of the work for you. But, because you learn the "long" way you're able to check the data for potential discrepancies. You also better understand the result.
Right. There is still very obvious value to teaching someone HOW to get to an end result, rather than simply giving them the result with no effort.
It would be like teaching kids that dividing 10 by 2 is 5, but not actually teaching them what division actually is. They'd know that 10/2=5. But they wouldn't know that it means that 2 goes into 10 five times, and vice versa. And obviously that is something they need to know.
Posted on 5/14/25 at 10:22 am to Samso
quote:
Does it explain how he “AI Proofed” his assignments?
It didn't say in the article from what I read.
Posted on 5/14/25 at 10:46 am to Bard
quote:
Yes, they are. The responsibility of the student is to learn how to retain knowledge and understand its place in their chosen field of study. Just getting something spit out to them by a computer doesn't achieve the learning experience (other than just plugging a question into a computer, which a trained dog can do).
Right? I've got an almost 15 year old that is getting ready for a drivers permit. For the last couple of years I have her navigate me home from different places, I randomly ask the names of the roads we are on, what direction we are heading in (which, as a woman, is not my strong suit
). I point out what roads run parallel to each other and roads we are coming up on. I quiz her on what direction neighboring towns and cities lay, I ask her what is in the direction we are driving towards (downtown, Target, hospital, Starbucks, mall, interstate, river, lake). All information she can get from navigation apps on her phone but I feel it is important for her to know the basics of where she lives without it. We've driven kids home from school who couldn't give us directions to their house without plugging it into navigation first. Posted on 5/14/25 at 10:49 am to BluegrassBelle
I've tried to use AI to write engineering reports for work. The result is always reports with unnecessary redundancy and VERY wordy.
With that said, it depends on your audience. If the audience is private industry, they just want the results presented clear and concise, no BS. If your audience is government like DOE, they value fluff.
With that said, it depends on your audience. If the audience is private industry, they just want the results presented clear and concise, no BS. If your audience is government like DOE, they value fluff.
Posted on 5/14/25 at 10:52 am to BluegrassBelle
That’s grounds for automatic failure in my book
Can’t even be creative when they cheat. Too damn lazy
Can’t even be creative when they cheat. Too damn lazy
Posted on 5/14/25 at 10:57 am to BluegrassBelle
AI will require a fundamental rethinking of educational approaches.
Posted on 5/14/25 at 11:07 am to GumboPot
quote:
Can some provide an example of how to do this other than in class hand written essays or problems?
You could utilize more oral assignments/presentations and ask questions in real time. Utilize assignments that require you to analyze and synthesize instead of just straight up recall. Break larger assignments into sections/parts where they have to show scaling of knowledge (think outline, first drafts, final drafts, etc). Use assignments that require multiple sources, current events, personalization. Use peer-review.
Sure, they can still use AI for some components of those assignments but they won't be able to use it for the whole assignment.
Posted on 5/14/25 at 11:12 am to CocomoLSU
quote:
To me, it's sort of like when calculators became ubiquitous. It doesn't mean that teaching long division is a bad thing, but it definitely doesn't carry the same weight it used to.
Good apology. When you know the basics, then you are much more likely to know when a computed value is incorrect. It also helps to build intuition of the system your are working. Without this intuition, you'll accept any answer without critically looking at the output.
Posted on 5/14/25 at 11:13 am to BluegrassBelle
Another way to AI proof assignments is to have them present the content and pressure test it via questioning live in class.
It will be very obvious very quickly who had AI-written papers.
It will be very obvious very quickly who had AI-written papers.
Posted on 5/14/25 at 11:14 am to Slingscode
quote:
When you know the basics, then you are much more likely to know when a computed value is incorrect. It also helps to build intuition of the system your are working. Without this intuition, you'll accept any answer without critically looking at the output.
Exactly. They're not being taught "how to think" and by that I don't mean indoctrination...I mean how to view and break down a particular problem or issue that we see countless times every day in our personal and professional lives. This next generation is going to be an interesting one when they reach the midpoint of their careers.
Posted on 5/14/25 at 11:18 am to 1BamaRTR
quote:
A lot of these kids are already struggling to read and understand things properly since they mostly consumed "brain rot" content as young children
Look at the rise of of companies like Grammerly.
“We’ll write your papers and memos for you because you're too incompetent to do it yourself “.
Posted on 5/14/25 at 11:23 am to Benne Wafer
quote:
Right? I've got an almost 15 year old that is getting ready for a drivers permit. For the last couple of years I have her navigate me home from different places, I randomly ask the names of the roads we are on, what direction we are heading in (which, as a woman, is not my strong suit ). I point out what roads run parallel to each other and roads we are coming up on. I quiz her on what direction neighboring towns and cities lay, I ask her what is in the direction we are driving towards (downtown, Target, hospital, Starbucks, mall, interstate, river, lake). All information she can get from navigation apps on her phone but I feel it is important for her to know the basics of where she lives without it. We've driven kids home from school who couldn't give us directions to their house without plugging it into navigation first.
This is great and totally agree. Be sure you teach her how to change a tire safely as well. She will thank you one day.
Posted on 5/14/25 at 11:24 am to BluegrassBelle
how do you "AI proof" an assignment? This sounds like more made up or exaggerated bullshite for clicks.
Posted on 5/14/25 at 11:24 am to BluegrassBelle
ChatGPT is great as a study supplement if you’re trying to make note cards or want to create a study guide quickly but the students that use it for every single assignment are screwed in life.
Page 2 of 3
Popular
Back to top
Follow TigerDroppings for LSU Football News