A committee of professors and students spent 5 months studying how AI changed learning on campus, and the findings read like a warning to every university on the planet.
⭕ Study groups are disappearing.
⭕ Office hours are emptying out.
⭕ Problem sets and take-home exams no longer prove anything, because AI can produce credible solutions to almost any written assignment in the undergraduate curriculum.
⭕ Students who lean on chatbots lose mastery and confidence, and some slip into what the report calls cognitive surrender, reaching for AI at the first hint of struggle.
The committee warns MIT professors against replacing undergraduate research assistants with AI agents just because they’re cheaper, because a university exists to grow people, not output.
Can we PLEASEEEE not shoehorn “TIL” into being just another repository for news stories? Imagine if I posted in here every day saying “oohhh today I learned what temperature it is outside today”. Clearly against the spirit of the community. Yes, its good information… So please put it where it belongs so that people can easily find it and easily find other things too, rather than mishmoshing all our posts into one unfilterable, uncategorizable blob
This will not change. This will continue because every one of those things hold short term benefits for individuals, long term be damned.
No study groups? Students will see that as a win.
No office hours? More convenient for students. Professors / admin save time or money.
Homework and tests No longer work means Students and professors wont have to do them, or no longer spend much time on them.
Students lose mastery and confidence means more reliant on AI, and there is no way this is lost on tech bros as a selling point.
The free market WILL NOT solve this problem. These are classic externalities and capitalism DGAF about problems that don’t show up on the cost ledger.
This has been going on for a while. Even before llms. When I was younger, I was actually educated as in trying new things and learning about them. Then came the proficiency test focus of leave every child behind. Then came in gutting classes and programs to only focus on tests and sports. Then came the charging money for sports. Then when I worked in education; it got worse. You would have kids that are 14 that can’t write two sentences; can’t write an email without auto-correct; struggled to read, it looked like someone smacked a keyboard to try and write something. And so on; it’s just the next level of going downwards.
It was sad to see that students with so much access to tech became so … I don’t even know.
When I was thirteen I was writing programs, using pretty much any multimedia software, beta testing games (how to get free games for 500), tweaking the lawnmower engine (to mow the lawn faster, that pesky governor lol), and so on.
So you’re right, this won’t change, maybe when it backfires it might. But this is creating a huge void, not even a brain drain anymore. There’s going to be a general malfunction to society. I mean it’s sad
It was sad to see that students with so much access to tech became so … I don’t even know.
Tech became too easy to use. In the 00s, you’d need to have an understanding of file paths. You’d need to know where a programmes files were located, to fix whatever broke
Eventually “auto repair” became the standard, then we moved onto handheld consoles (smartphones and tablets). Which relied on app stores, that already confirmed the app wouldn’t destroy anything. So all these annoying to deal with issues just… Vanished. But so did skill, because you never needed the skill to solve a problem in the first place
The skills simply change over time. It’s great that you could muck about and maybe fix something back in the day, but does that same skill apply to modern architecture? No. The true skill you learned what your ability to be curious, to constantly learn and improve, and use those skills across a wide variety of problems, not just simple tech issues. Those are the skills we should be teaching the next and future generation, rather than what the classic education paradigm has been, which is memorization and how to show up on time, to make good factory workers, which are existing less and less in the modern world. Does AI help or harm that? It’s just another tool that can be used in good or bad ways, just like the calculator.
but does that same skill apply to modern architecture
Architecture hasn’t changed all that much. It’s just more hidden
AI harms curiosity
I disagree. For me it’s helped with curiosity as it lets me explore topics and problem solving I haven’t been able to because I don’t know how to fully finish something. Oh there is this great github project I would like to run, but it’s linux only, is it possible to integrate this on my mac and add the following to it. Sure enough, LLM helps get you through it, you document the process, and boom now you have learned something new.
In a way you’re right but a lot of this comes at a cost. As in people are not used to being bored anymore. Not being bored leads to less creative thinking and procrastination. Internet is out, well I guess I’m working on the motorcycle instead of sitting at the computer.
How do you get the connections to beta test games?
Now, you have to know someone to be referenced if you want to get paid. If you want play games for free, you can always apply for closed or open betas; at lot of times stress tests are needed as well as hardware tests. Then again “early access” has become a bit of a plague to that.
You also get the perk of not reporting glitches. I don’t know if you’ve played World of Warcraft, there was a fun reason they implemented a global cooldown for skills. One reason was you could write a macro to swap weapons and do a backstab or ambush with the slowest blunt weapon possible. Oh the joys of 4k to 8k crits. It was quite a bit of a “solo leveling” dungeon experience. I mean if you think about it, getting
backstabbedrearended with a big blunt mace probably would be a more serious crit than a tiny dagger.
Has the MIT published a report on how it’ll try to get AI to commit suicide over copyright infringement?
even a random idiot could have called that last point.
the ground work for this was laid when search engines were useful. People would stop thinking, they’d just break out the phone and search for the solution to any problem… and they never retained the information, because the brain isnt designed to retain information in situations where you can pull the magic slab out of your pocket to do 3 seconds of problem solving for you.
and its only gotten worse with AI. There are people who literally cant even function without an AI telling them what to do anymore, Its taken over their entire lives and doing almost everything for them except telling them when and how to go to the bathroom… and I wager it wont be long before we start having people pissing and shitting themselves because their AI didnt tell them they should go to the bathroom.
They don’t want people to think of ways to rise against them.
yep.
the terrifying consequences of AI use are the final stages of decades of attacks on education… creating the perfect being. A complete, non-functional idiot that believes everything its told without any thought or consideration, and completely non-functional without their masters grace (the AI)
God damn, I hate how accurate this is.
we are quite literally going down the idiocracy pipeline. i sure hope some secret military operation to send 2 completely average people into the future 500 years is taking place
secret military operation
operation electrolytes.
deleted by creator
TIL is supposed to be like an interesting or suprising fact. or an easy overlooked lifehack.
this is like randomly picking a research paper and having chatgpt summarize it, then post it as TIL .
additionally, do we even know that this is a fact? its really just the outcome of their own case study. it probably says something about the intersection of education and technology at the current moment. but that could all change in less than a year.Thanks for posting this. It bothered me too but I wasn’t sure how to put it into words. It’s almost as bad as the soapbox rants that get posted to the asklemmy comms with “What do you think?” at the end.
I had Claude read the results. Sounds like the changes are all positive and nobody should rethink their investments or effect on society at all
I just realised why I find AI enthusiasts so repulsive. I was already aware of how rude it is for someone to share the output from an LLM. Someone sharing an opinion they formed based on such an interaction, I realised is as repugnant, if not more.
Or to put it in a different way: “I couldn’t form my own opinion on the matter. So I shook a magic 8 cube a bunch of times, and the ‘conversation’ we had, is something I think is worth while for others to hear”.
If it’s a month old, it’s probably news vs TIL.
Notice that these are also all ways to eliminate the few remaining places where people actually first meet in person, realize they have things in common, form groups/relationships, organize towards common goals…
It’s like they just want two people, sitting in separate rooms, depressed and constantly delivered targeted messages they control about how the other person, in the other room, hates and judges them.
Kinda sounds like a plot.
How to solve this, then, from these bullet points:
Hire more grad assistants. Get them to administer one-on-one oral defenses —they’ll only need 10 minutes or less — of the undergraduates’ take-home exams. Watch the undergrads fucking crumple and get washed out. Those who are left are the ones who’d use office hours and study groups anyway.
Agreed except the office hours/study groups. I worked full time at uni and could not go to these as often as I wanted.
My partner was a grad student in Italy and I liked how they administered an oral exam for each student. Pretty easy to separate wheat from chaff that way I think, and quickly
Why not just start enforcing Anti-AI? Underperformers would still be tempted to use AI and then flunk themselves when they otherwise could have succeeded.
They do, but it’s an arms race and assessors have limited time. And if you get it wrong (and they didn’t use AI) then you’ve put an innocent person through a process that can be rather degrading.
So it ends up being that only the most egeregious cases get addressed and the smarter cheats get through.
Cheating in college was already an arms race, yet nobody asked for it to become the norm…
Appreciate the summary, although I don’t even like holding office hours so I don’t mind that. The rest are depressing, though.
That said… my own experience doesn’t match up on the take-homes being useless. My God the way answers math problems are so bizarrely incorrect that I’ve given more 0s than I have in my entire fucking life.
At least, upside, my holistic approach with a lot of discussion, projects and oral components to assignments are really shining. Helps that my classes are only 25 max though- I can’t imagine how bad 80-300 classes are.
I hate this so much
AI will not make people more productive
I dunno, for me at least it’s less about ‘AI’ making me more productive, but rather automation that absolutely makes me more productive. Ai certainly makes the process of getting automations in place a whole lot quicker. Any time I do something repeatedly that I don’t want to do, it’s the question of ‘how can I automate this process’, and that frees up time for other things, and ensures that what I do is more accurate. That’s not something new because of ‘AI’, but rather python has been the beacon for that for many years, with tons of books on the topic.
AI is like the microwave. Sure it a huge time saver to eat a microwave meal but that doesn’t make a good cook obsolete.
Absolutely agree. But the car certainly made horses overall obsolete when it comes to standard transportation.
AI isn’t a car and will never be
The people who are pushing that idea are the ones who will lose billions when it all comes down.
I don’t think the Chinese will lose anything. They are investing knowing AI will be the future and making it cheap and accessible. Same thing they are doing with EVs and green technology.
Unless they persuade enough gullible people (or corrupt governments) to eat the losses for them. Which is what they are trying.
Creating efficient automations is the basic must-have-skill of a good engineer. If you outsource that to a slop bot, you surrender to being a replacable mediocre button pusher. And your slop automations will be shit.
I disagree. Certainly for systems architecture you may need a ‘good’ engineer, but you really don’t need to hire someone to automate the routine clicks you are doing every day. Oh look I have to pull and move the same data everyday, can’t I write a python script for that? Yes. And for many who are not programmers, an LLM can easily do that for you, and works just as well as that junior dev who is overloaded and can’t be bothered to do that for you.
You have the right disagree, without being right about it. That “junior dev” mindset is PR speak for “reasons to pay them less”. Actual developers have acquired skills as hobbyists long before entering a job, and will self-train continuously on the job. That’s how you get senior developers. Using a slop bot to generate code is something that 1) an actually skilled dev wouldn’t do because they have zero time gain from it, and 2) it will mean you keep dumb people dumb, never having them learn skills as time passes. Non-deterministic coding assistants are for gullible people who want to license-launder other people’s code and/or wouldn’t ever be able to create any proper software. And yes, that includes people who suffer from brain deterioration after having once been decent devs.
Perhaps. Every senior dev I have talked to absolutely is using AI right now. Slop code is relative.
Spoiler: those are not senior devs, whether they believe so or not.
The conclusion of the report looks like they are all in on saturating their university with ai use
Silver lining is that even before AI, courses were bloated with busywork. Though it likely won’t happen, there’s a possibility that trimming down on coursework would save a lot of people a lot of time.
Save time on learning? Isn’t that where you should take time?
no, silly, learning is a waste of time.
half the people on lemmy basically think education is stupid and pointless unless it immediately and directly needs to employment. they’d rather we stick 13 year olds in internships.
If they only have 5 minutes of homework tonight, then that should be OK.
I publish book reviews on business books, the interactions have almost dropped to zero as have people in those channels. I imagine it’s related.
I try not to have discussions about AI online that are too lengthy, because I know no one really convinces anyone else, but every time I see someone talk about how they use local open weight models and how that makes their AI use ethical I think of scenarios exactly like this. That AI got the information from somewhere and that book or forum post or think piece is not getting the buys/engagement/ad revenue because the AI basically stole it. People aren’t joining communities and adding to a public knowledge base. They are just quietly solving their issue alone without contributing to the ecosystem that helped fix the issue and the ecosystems that sprung up around those.
It feels like a loss of so much. A loss of interest in research, but also of community, of a desire to share. I remember being on forums and knowing that if user X responded it was going to be the most thorough, interesting read. I remember following specific users because even if their post wasn’t directly relevant to me at the moment, I wanted to learn all I could from them. Not to be an old man screaming at a cloud, but I think AI is stripping a lot away from the younger generation that they don’t even know to miss, and I think that’d be true even if AI always provided the most engaging, in depth, accurate responses you could find on the internet.
I feel every word you typed. 💔
I feel for you. I’m sure for authors, publishers, and consumers that such books and reviews have lost all credibility if sloppers can generate fakes with ease. It’s difficult to know whats real and can be trusted versus what’s just wasting our time.
Without a doubt.
And people who don’t know any better will believe what AI says. I see it at work all the time. People producing pages and pages of slop.
People join calls, output AI notes that clearly show they weren’t paying attention at all.
Sincere question: Can anyone explain to me where the youths are finding LLMs that write half-decent prose?
2-3 years ago, I would’ve bet money that the future was gonna be bluebook exams for every subject, but clearly I was wrong. I don’t get how you successfully cheat with ChatGPT in a productive way if you’re writing on paper with the teacher watching, call me old fashioned.
Does it matter if it can’t write half-decent prose if none of your classmates can either and the system just adapts to that to avoid failing 70% of students.
