You must log in or # to comment.
This is kind of the argument I have with my coworkers about AI. One in particular loves using ChatGPT to generate all the bullshit memos, justifications, and accomplishments we have to produce, however I have to point out that those things had no value to begin with. AI is good at doing things that don’t need to be done.
Always has been…
I had Latin in school, and the teacher was a bit senile already. He started making the rounds to check our homework - which I had not done. So I just hastily scribbled some lines of gibberish into my notebook. He came to my place, looked at my notes, and scolded me for my bad writing. And signed it off as done…
Non scholae sed vitae discimus!
For my vitae, Latin has been rather useless in the last 40+ years since I got rid of it.
It was just the choice of the lesser evil back then: French with an asshole teacher or Latin with a demented one. At least the teacher I had at the end let me off with the equivalent of a D instead of an F.
Illigitimus non carborundum!
Lorem ipsum!
E pluribus unum!
Ecce, Romani!
Tepus fugit.
Cornelia et Flavia non quod ita vero sum.
Romanes eunt domus.
Semper fidelis tyrannosaurus
Homework is not to measure anything is to force you to make the work so your brain develops new connections.
Using AI for homework or study is like bringing a forklift to the gym.
I suspect a large part of the problem is that, at least in my experience, this is not at all explained to students and often the teachers themselves do it as “something that everybody does” rather than with the understanding that there is an underlying purposed that homework is meant to serve (which sould inform the when, what, how and how much of homework).
If see a similar kind of problem in my area (Software Development) all the time - people doing certain things because “they’re good practices”, “it’s what you’re supposed to do” or even “that’s what everybody does” without at all understanding the underlying reasons for doing it (and, more importantly, when to do it, when not to do it and how best to do it), which is why for example nowadays you have countless of “Agile” teams that are doing wrong or unecessary (in their context) parts of it whilst not doing the parts that the should do just doing things they think they’re supposed to do it but doing them incorrectly since they don’t get what those things are supposed to achieve and how.
Mind you, some people are just lazy, so some students will just “optimize away” homework with whatever tools they have to do so, even when knowing the purpose of homework and when being given the most learning-enhancing homework possible.
Being able to use exercise equipment with a forklift would be pretty cool tho
To be fair, you’d have to develop mad forklift skills to do anything outside of deadlifts.
That sounds way more fun
Going to a gym to pose as doing exercise while not doing it? Yeah fun.
Bringing a forklift to the gym and trying to pick up all the different shapped weights.
I just spent the afternoon driving 79" helical Piles with a dingo for a solar array, the n grated the land, and since I had an extra hour with the machine , helped landscaping crew on site dig out a trench to plant hedges.
Easily the most fun ive had at work in a month.
It’d be a different sort of exercise, but it would be interesting as a means of learning how to control a forklift.
Yar well, in this situation school is for the muscles. Might be fun but it won’t make you stronger
Sure, but within the simile that’s sitting at the back of the class and laughing with your friends instead of engaging with the lesson. Fun but you didn’t gain anything and got in the way of people actually trying.
No the forklift thing
deleted by creator
Synthesis: Learning is a mistake, return to ignorance, just figure stuff.
Think about it, do you really think it’s a coincidence that all major religions reject asking questions??
Typically, when I want to demonstrate the full breadth of my mastery of the ethos/logos complex to an authority figure, I mostly talk about orthogonal relativity and how it is möbiation of entanglement within our topological matrix that gives rise to higher order thinking, like ordering breadsticks at Arby’s.
I mean this is exactly what all the kids who told me they loooved essay questions always said they liked about them.
It’s too bad, now we’ll have to replace them with speaking in front of the class without notes.
I would love that!
I wouldn’t mind. Some would struggle.
I taught a class of eight students who liked each other and got along well (but weren’t dating each other) for twenty hours a week for six weeks. They were all still terrified to give presentations at the end, even though it was a language course and they all already spoke a lot in class. I don’t know how I could make it easier for them, because that’s just about the lowest stress scenario I can set up. I’m not judging them for it, because I had a couple moments of panic in front of them as well, sometimes it happens, but I’d like to make it less stressful. I could just have them do it every day to make it routine, but if some of them are really affected by it, that might just make them dread the class.
Public speaking is a dying art, with the potential to change the world. People just aren’t given enough exposure in the right circumstances anymore. All we need is a place where it’s safe to fail, and all that nervous energy becomes a vibrant excitement with no comparison.
same, they all love cheating, before AI they would mostly just copy paste eachother’s work
and me the alone kid who actually wanted and cared about learning would be drown out in busy work to be done, get bad grades even though I did well in tests and exercises, simply because I failed to do the annoying boring homeowrk of writting an essay that we all know the teacher is not reading, that they just made thjs assigment to give students free grade points because they know that if they didn’t half of the students would fail the class, they would go lenghts about how they were being nice by doing this, and yet screwing over me would never be mentioned, one of the few students that ever displayed any form of curiosity in class
i can tell you from experience that teachers are using it too lol, to create the homework
one of my university professors admitted that he used copilot to create some questions for our exams of that class 🙃
Then you had a bad professor and the homework was useless. What class was it?
software engineering processes (the name is in french, this is my best attempt at a translation)
but many of my classes this semester had ai slop in them lol, one of them had nonsensical ai generated images in the class notes, in another one the teacher used cursor as his IDE to demonstrate stuff…
my classmates are all in on it too, naturally. for one project, one of my teammates announced that he’d already done most of the work! wow, so cool, and so early too! in hindsight i should’ve seen it coming… basically the whole thing was vibecoded and i only noticed at the end when it was time to do minor adjustments (such as fixing major features that were not working lol)
AI is terrible in university. Because there is minimal effort on the students to produce it, and the hours I waste marking it is wasted. Pointing out the errors produces no value since The students didn’t go through the process in the first place, and the machine isn’t listening.
I see the same thing in my day job. Analysts produce effortless reams of bullshit that technical experts like myself have to wade through and proof read. We are seen as the barrier, but the more AI is used the longer the review takes because there was no quality control on the generation of the material.
Would this be better than recycling the same 5yrs worth of material?
What’s wrong with recycling material? If it’s decent, it’s still gonna be useful for a new class…
When your exams are all the same questions every year just in different orders and all your exercises are on obsolete software, yeah, it’s a problem. Even worse is when the prof doesn’t even look at your work and just gives a mark based on his feels that day. I did extremely well in that class, but all I learned was I forgot something in an autoexec.bat file, I didn’t realize it went 1 line past the bottom of the screen with a silly broken command
Autoexec.bat… you’re over fifty :p so am I though :(
I tried it for my class, and the questions they come up with is boring, repetitive, and generic.
I feel very sorry for you that you need to endure that.
Yea, I saw LLMs try to ask me questions to test my knowledge (which I didn’t request) and they were so bad, I was almost feeling second-hand embarassement. They would ask a rather obvious question, and provide a tip that was almost the answer itself.
As a public school teacher, part of our training on AI is to use it in our classrooms and to talk about using it, like you would in making a citation.
I do use it to construct rote tasks, especially materials for vocabulary practice which are great but aren’t worth my energy to spend much time on. I always tell the kids when I’ve used AI to create sentences, etc. I think it’s great for them to see usage modelled responsibly.
It would be a disadvantage to deny kids usage altogether, and prompting AI should be explicitly taught as a skill set. Cheating is definitely an issue, but more and more teachers are moving away from rampant computer usage in class and thinking actively about how to forestall such cheating.
Yes, but, if you can’t produce plausible bullshit on your own you’re not going to get very far.
if 90% of jobs can be replaced by AI bullshit, why tf do we even need to work? Why shape society so you must do shitty non existent jobs to survive?
May I introduce you to the work of David Graeber on this exact subject? Bullshit Jobs
I forgot that he also wrote that. I read “the dawn of everything” which is one of the most mind-blowing revisions of history I’ve ever seen. (In a good way, for once)
I recently finished it and it completely changed how I see society. It’s the intellectual equivalent of an acid trip, with how much it breaks down and challenges things you’ve just always assumed.
I also recently read Debt and The Utopia of Rules. I highly recommend people read everything of his.
Thanks for link!
Automation should enable UBI, in a decent society. But since our current leaders will never implement UBI, automation is a crime against the working class who need jobs to survive.
Correct answer, we are already in the UBI realm of not needing enough workers. The question is where the world goes from here. I’m currently betting on some dystopian societal structure we’re 90% of people are fucked. So glad I’m older without kids.
if 90% of jobs can be replaced by AI bullshit
Well, to start, they can’t.
At least not by these LLMs that keep getting touted as AI.
You are right because 90% of workers are not middle management!
Though this doesn’t seem like a valid application of TAS, I’ll bite.
A calculator can do your math homework for you, but that doesn’t mean the homework is useless. It means that if you don’t follow the rules the work is useless. Just because something with no understanding can complete a task doesn’t mean a human with no understanding can. It is built to test a human, not a computer, and it generally does so well enough provided the human is the one doing the work.
Yeah it’s built to serve as practice and to leave you with the questions you need to ask the next day.
I get it. I hated homework in school and generally didn’t do it. And a fair bit of it was busywork. But when I got to college I learned the hard way that if I didn’t do my homework, I wouldn’t really have the skills. It’s easy to sit through classes daydreaming or even listening, and just not really understand how to do a thing.
How is anyone using ChatGPT in an exam? Surely exams are taken in controlled conditions?
deleted by creator
That is absolutely wild.
When I was at uni (2016-2019), you had to leave all of your belongings behind in a designated area, and only carry a clear/transparent container with your pen(s) to the exam room. You could optionally bring a clear water bottle (label removed.)
If you ask to use the bathroom, there is no opportunity to go back to your belongings.
I’ve gotten my period by surprise during an exam, I’m glad I didn’t go to a university like that.
Most of my exams have been online with no proctoring… Hell most of them are meant to be open note I think. That or my class just does a final project or paper rather than an exam. I’ve only had a few in person on paper final exams, and one online proctored exam where you have to scan your surroundings before testing.
I hate AI so I’d never use that to cheat (plus that shit is obvious and often wrong anyway), I just look at my own notes because I wrote them for a fucking reason. But like most of my exams or papers could’ve been done with AI I guess. Or like someone said go to the bathroom with your phone, as I’ve never had to hand over my phone before an in-person exam.
Honestly highschool was more serious about devices when I took state tests/finals and the ACT, they either took it or it had to be in a bag completely turned off, and if you did take a bathroom break (which was HIGHLY discouraged and I think only available during certain times), you had to hand over your phone.
My students can use a translator for every interaction they have forever, or they can do the homework and try to learn the language. Three guesses as to what’s actually easier in the end. Just because it can be done by a computer doesn’t mean there’s no reason for a human to learn it.
I don’t think the point was anti-learning so much as it was anti-measurement
I hear you, but language class is actually a decent example. I met a language requirement of 2 years in high school. Then I turned around and completed the required two semesters in college, of the same language, starting over at the beginning. I don’t know anything but the most elementary French.
In high school I did all the homework and felt lots of stress and really struggled. In college, I google translated my way through all the homework, phoned it in and same result, way less time and energy. So using the computer was actually easier in the end.
I’m probably never making it to France. Maybe not even Canada. You could argue it was still good for my brain to stretch in that way, but it was just never going to click for me and was a requirement regardless of my actual educational needs.
My single semester of sign language I actually picked up on decently and it tickled my brain in all kinds of ways, and I actually used it at work once to help a dude. No one’s ever come into any of my workplaces speaking French (at least not yet)
That’s a totally different situation, to be fair. I’m teaching German to recent immigrants to Germany. They will be much, much better off if they know the language.
The thing that this person seems to have either forgotten, or not understood in the first place, is that homework, and your education in general, is not for the teacher, it’s for you. If you choose to cheat your way through you will gain less than if you actually put the work in yourself. This gets more important the further through education you go.
Probably the best outcome for an essay question is if you discuss it with your friends, all share your understanding of the subject, then write it up individually, incorporating anything new you learned from your discussions.
This does come with the issue that those who cheat could end up getting good grades if there is no ‘live’ check of their understanding, either through closed book exams, class tests, group discussions, or similar.
Indeed, it’s not that LLM is totally uncorrelated to factual basis, it is correlated well enough for most educational fodder that is supremely well trodden and doesn’t seek surprises. It’s practice with backgrounds the teachers can actually already know and credibly provide feedback on. It’s so well known that any teacher knows, so it’s correlated with AI output that has been trained on verbatim prompts and essays that are on the exact same topic, and so the narrative correlated with the prompt is just very likely to also correlate to the facts of relevance.
Just like a pocket calculator can make short work of elementary school math. It’s not that we expect those kids to do some crazy novel stuff that calculators can’t do, it’s just that they need to operate in a context to actually illustrate they have the foundational understanding.
your education in general, is not for the teacher, it’s for you
Most teachers don’t understand that
You were so close. Right up until you mentioned closed book exams. How many times in your job have you ever not been allowed to look something up? The reality is that closed book exams only test your memorization capabilities and some of us don’t memorize shit very well. The best teachers allow open book exams because if you know where the information is then you can find it in time but if you don’t know where the information is you’re not gonna be able to take the test in the time given.
How many times in your job have you ever not been allowed to look something up?
It’s rare, but not unheard of, that I can’t look things up, but the point of closed book exams is is to demonstrate that you know the subject well enough that you don’t need to look things up. Obviously, exactly what this entails is going to vary depending on the level if the exam. If it’s testing foundational knowledge, then it should all be in your head, if it’s more advanced, a crib sheet with key facts (say certain more complex, but necessary, equations for a non maths subject, or similar support prompts).
If you’re working, you can’t be stopping every few minutes to look up basic information. A computer programmer who has to keep looking up the syntax of their language, or basic algorithms, for example, won’t get very far.
I’ve been a computer programmer for 15 years, Not once have I ever been in such a time crunch that I couldn’t double check something.
There’s a big difference between double checking things occasionally, and needing to look up fundamental things. I’ve been a sysadmin for, well, a long time, and there have been many occasions where I’ve been pulled in to rescue a situation and had to rely on what I knew, without being able to refer to other material, either due to intense time pressure, or enough being down that there isn’t anything to reference.
Besides, exams should be testing your core understanding of the subject, the sort of knowledge everything else is built on, and your ability to apply that knowledge in different scenarios. Practical tests, are better suited to assessing how you use novel information, do more advanced things, and handle reference material. Maybe we’re talking about the same sort of thing in different ways?
I think both closed book exams, and practical tests/discussions and the like have an important role to play in assessing your performance, whether a teacher sets them or you challenge yourself.
The algorithms one in particular is a bugbear of mine, because if you don’t already know the computational complexities of the operations on the common basic data structures then no way are you taking the time to look them all up each and every time you declare one. And yet one of the bitchwhiniest complaints I frequently see online about coding interviews is how dare hiring managers ask you to prove you understood CS201 Data Structures & Algorithms…
This is one of those “you won’t need it, but the smart kids will” things.
I’ve used Pythagoras at least four times in my professional life, and not one of those occasions was the problem presented to me in the form of a neat “Here is a right-angled triangle; what is the length of this hypotenuse?” that I could have then looked up in response if I didn’t already know.
Right, but that’s not the point I’m trying to make. The point I’m trying to make is during any of those times was there a point in which you couldn’t look up what the Pythagorean formula was?
I feel like you haven’t understood my point either. If I didn’t already know Pythagoras, I wouldn’t have known that it was a valid solution to the problem I had, therefore I wouldn’t have known to look it up.
No, that was literally my point. That it’s more important to know how to find information and where the correct information is then it is to memorize the exact information. Knowing Pythagoras and having the formula memorized are two very different things especially for things that get more complicated than Pythagoras.
I’m a language teacher. It happens all the time that I want to say something and can’t look it up. Think about the last time you were in a group of six or more people: the conversation doesn’t stay on one topic long enough to look up a word before responding to someone’s comment. I also wouldn’t want to take a moment to double check a word and hold up the whole line when the cashier at the grocery store asks me a question.
I think that situation is going to entirely depend on your friends. But your point is taken, however I still feel like you’re describing the same paradigm I was in which you know enough about your subject to look up the word if you wanted to or needed to but you don’t need to because you know enough about your subject to say other words that mean the same thing. My point was that it’s more important to know how to find the information than it is to simply have the information. Just being given the information is literally vibecoding.
Depends, with homework, the goal is to learn, but with tests, it’s already less clear, a good teacher needs to know the level of it’s class in order to curate it’s teaching to their understanding. Moreover, some formation recruit based on grades and in this context, the point op is making is very relevant
When I am struggling with Language Arts and excelling at Math, the extra time spent on unmastered Language homework benefits me. The extra time spent on Math serves only to waste my time and distract me from the Language work that needs my central focus.
The intention of homework is to benefit the student. The reality is that most students would be best served by skipping the homework and getting some exercise instead.
In the gradebook, the unit test score should be the floor for all homework scores in that unit. Score a 95% on the test, and even if you turn in zero homework, a 95% should be recorded for each assignment. S
Turn in 95% on each and every homework assignment, but score a 65% on the test? You keep your 95% scores showing you put in the work, even though you clearly haven’t mastered the material.
Turn in no homework, and score a 65% on the test? Every homework assignment is now a 65%.
With this system, doing the homework can only help you. Skipping the homework is risky, but allowable.
Why would you think that ChatGPT has no understanding of facts or semantics? I think whoever wrote that has different criteria between intelligence and AI when it comes to understanding.
I can’t speak for the original author, but I suspect they probably think that because it is the truth.
It’s not that I think AI is better than you do. It’s that people who say things like you do are setting different standards for AI and human intelligence, but you don’t explain why you think it’s okay or reasonable to do so.
I mean, the provable lack of semantic comprehension is literally the mechanism for a subset of academically rigorous papers on LLM attacks.
There are plenty of things you can debate in the body of discourse around LLMs, this just flatly isn’t one of them.
Those academically rigorous papers prove a complete lack of semantic comprehension? Exactly zero semantic comprehension? Because that is the claim here. “NO UNDERSTANDING OF FACTS OR SEMANTICS.”
It’s strange because all of the papers I see deal with semantic error rates, which is a completely different claim. Now, then, how would you say that you are personally doing with regard to semantics?
Me? Personally? Tough to say, you’re entering into the realm of philosophy at that point… but it’s a more fruitful avenue of argument… even though I know you were just trying to be a dick.
I genuinely think it would be an easier task to argue human minds don’t even qualify and then to argue LLM equivalency than it would be to argue an LM has semantic comprehension.
Okay, you just said:
I genuinely think it would be an easier task to argue human minds don’t even qualify and then to argue LLM equivalency than it would be to argue an LM has semantic comprehension.
Now, don’t just breeze through this, actually read what I said in my previous comments:
I think whoever wrote that has different criteria between intelligence and AI when it comes to understanding.
and
people […] are setting different standards for AI and human intelligence
If you think what you said you thought, then why would you argue against me in the first place?
It seems to me like in your first comment, you were steel-manning the original argument by completely ignoring the obviously completely untrue part about “facts”, and pretending that it didn’t say there was “no”, as in “zero” semantic understanding.
Meanwhile, you’re straw-manning my comment, ignoring the fact that all I said was saying we need to compare AI and human intelligence by the same standards. So here I have an apparently human intelligence that is arguing against a meaning that I was not conveying, meanwhile supporting a different meaning that the OP wasn’t conveying.
You can’t steel man one person’s argument just because you support it, and then straw man the opposition’s argument just because you don’t support it.
What you conveyed in that comment was that your understanding of the meanings in the argument up to that point was completely fucked. It’s the perfect example, so of course, I’m going to reference it, even if it makes you think I was just being a dick.
I’m sorry for glossing over it, you’re right. You said a few times that you feel like humans don’t measure “intelligence” fairly towards machines.
You’re right. We probably don’t do this fairly towards animals. We probably don’t even do this fairly to other humans.
It’s a fair mirror to hold up.
Acknowledging that we’re susceptible to bias isn’t evidence that a conclusion is wrong, though. Some blondes actually ARE dumb.
This, for example:
I read the paper until I got to what I consider an indefensible statement about the relevant point. It was early in the paper.
success on benchmarks only demonstrates potemkin understanding: the illusion of understanding driven by answers irreconcilable with how any human would interpret a concept.
This demonstrates a different kind of understanding, not “no understanding,” or “the illusion of understanding.”
The idea that the understanding of a normal adult human (I’m being kind and overlooking their giant mistake of using weasel words by saying “any” human) is the only non-illusory type of understanding is indefensible. The best chess computers make moves that no top grandmaster could make in a game, but that can’t mean that the computer’s understanding is an illusion. It’s simply different.
The authors of that paper have made the mistake of making statements outside of their area of expertise. They want to show specific problems with LLMs in their subject area of computer science, but have made incorrect statements in the area of philosophy in an effort to do so.
You should not take their philosophical conclusions as the takeaway from this computer science paper. But I appreciate that you took the time to actually find something relevant.
Did you actually read their example tests? Are you saying that you can have a valid and useful definition of intelligence that includes those kind of mistakes?
Okay. It seems like I’m going to have to make this extremely simple.
Do you think a dog has intelligence or understanding? If so, if you ask a dog to write a poem, and the dog fails, does that mean the dog has zero intelligence and no understanding of anything? Or could you say that the dog simply has a non-human understanding of things?
What if you read that entire research paper to a chimpanzee who knew sign language, and you tried to make a point about philosophy to the chimp that had nothing to do with the tests, and was completely localized to a part of the paper that you quoted? And what if the chimp then tried to bluster and ask whether you even understood the example tests? Would that mean the chimpanzee has zero intelligence and zero understanding of anything? I would argue not. The chimp simply has a different understanding because it doesn’t really understand the topic.
That’s not what you claimed in your original comment. You said
“I think whoever wrote that has different criteria between intelligence and AI when it comes to understanding”
The paper I posted is literally applying the same criteria to humans and LLMs.
It seems like you are the one applying different criteria?
You’re neglecting that my comment was in response to the proposition that “ChatGPT has no understanding of facts or semantics.” If your paper isn’t probative to that point, which I’ve already demonstrated it misconstrues out of apparent ignorance of a wealth of existing philosophical papers, then it’s about as relevant as comparing human body temperature with processor temperatures. “I’m measuring them both with thermometers, so therefore I am using the same criteria for both humans and computers. Game. Set. Match.”
You’re caught up in some sort of weird pedantry while ignoring the overall meaning. In other words, you’re misunderstanding the semantics of this argument.
Link one or two of those papers?
Yeah but for real proof of understanding has always been a problem. One of the reasons generated text is so pernicious is because even in everyday life, it outwardly mimics all the signals of care and cognition.
To the main point: much smaller class sizes is how you solve this problem.
