AI use leads UC Berkeley math professor to see 'no point' in grading homework

>This is common practice among tenured professors with grant money to spend and graduate students beholden to them (I’m not making a moral statement, just noting this is common practice).

No, that's not common at all, nor how the system works. Use of graders/TAs are on an institution basis.

"Grant money" has absolutely nothing to do with it. Grant money is earmarked for specific research and cannot be spent for teaching purposes. When grant money is used to hire graduate students, they are research assistants and must work on the research project which the grant money was given for.

TAs/graders come from separate tuition budgets. Tuition dollars are much more preferred anyway because they're stable and faculty do not have to work for them, they show up based on enrollment.

The fact you have TAs/graders at all is by institution. Large R1 institutions have grad students and undergraduate population to fill these roles. In these institutions, 1 minimally sized undergrad class = at least 1 TA, with additional TAs and graders after some enrollment threshold.

Tenure has absolutely nothing to do with it. Adjunct faculty get TAs/graders all the same.

>The irony here is perhaps missed by many. He didn’t grade homework in the first >place. He paid assistants to grade. This is common practice among tenured >professors with grant money to spend and graduate students beholden to them >(I’m not making a moral statement, just noting this is common practice).

Grading is part of what it means to be a TA, and being a TA is not some awful exploitation, it's what gives graduate students living money (I was a TA in the 1990s). It makes no sense (except maybe in a small liberal arts school with no graduate students) to have professors grade.

One of my freshman math teachers gave grades out for 3 tests and none on homework. He also told us that if we didnt take the first 2 tests, it wouldn't count against us and only the grade on our final would count. Along with that he had extensive open office hours.

It was the math class where I learned the most as well as I wasn't doing rote work and was able to talk to him whenever I had an actual question.

I dont like AI destroying some of our systems but if we return back to the sort of learning that college used to be where it was self directed with guidance from professors, I will count that as a win over the factory worker production system that high school was and universities were starting to ape.

I teach CS at Berkeley, as a lecturer. i.e., I'm contingent and on a contract. And don't really have grant money—though, y'know I actually try to supplement my income so I can live in the Bay Area. I spent years working in EdTech as well.

Others have already say why TA's (and graders) often do the grading. That's also not the same as the prof not paying attention to the work students do.

But I do want folks to think about the time faculty are actually allocated to teach courses. I have a 2 unit course I teach, ~70 students this term, and 2 hours of class. I'm paid for maybe 8 hours per week for all the prep class time, planning. Doesn't actually leave much time at all for grading. There are definitely qualified students who could be TAs who would give the students much more time and feedback than I am able...

For other courses, I have a bit more time allocated, but hundreds of students and still one of me. The only way students get feedback is when TAs grade.

Anyway, that's just the practical bits...

A lot of faculty are moving to 100% proctored assessment, and I think it sucks. For me, reading at least some of the student responses and seeing the data are helpful. Faculty do want to engage with their students and are similarly demoralized when everything is reduced to a transaction. It's hard not to be nihilist about it. Our own jobs, even at a public University are just capitalistic transactions with fewer profits. Students want the degree because it's the only thing they can hold on to as a token of hope for the next few years. Both students and faculty can fall into patterns of really low trust, and it's hard to be the 'outlier' trying to trust students when most of your colleagues say people don't do the work.

I don't actually have great solutions for building intrinsic motivation in students, except talking to them and minimizing the extent to which everything is a transaction. But I'm required to assign grades, still.

Many of the 3rd-tier institutions are going to go bankrupt anyway. Demographic changes had already made that inevitable but AI has accelerated the timeline. If they want to survive at all then they'll need to reinvent themselves as vocational training facilities or senior assisted living or something. This is going to be rough for the faculty, and will make the process of finding a tenure track teaching position even more brutal.

> The irony here is perhaps missed by many. He didn’t grade homework in the first place. He paid assistants to grade.

Whether he graded it, or his assistants graded it, I’m not really sure how that matters at all to the point.

> They are going to college to increase their earning potential, and that’s about it.

Human Interviewer: Congratulations on making it to the 12th interview!

Candidate: Hey, I'm honored to be talk to another human for once!

Human Interviewer: That's funny. And true! Now, it says here you're a recent college grad student.

Candidate: That's right. Had to increase that sweet, sweet earning potential.

Human Interviewer: Also funny, also true! Honestly, what I learned in Intro to Game Theory alone has probably increased my earning potential by 15, maybe even 20%.

Candidate: Oh man, I remember that class! My professor gave up on grading homework that semester because of ChatGPT, and that weighted the midterm and final so much. I never did so much plugging and chugging in my life!

Human Interviewer: That's great. Well, that's about it. Hey, it was a pleasure meeting you. We'll be in touch.

Candidate: ???

The situation in American academia is something long mirrored in colleges in other countries.

You went to college because it was the only way to get a job that paid well.

When other vocations stopped having the ability to ensure people could enjoy social mobility, people focused on the only pathways left.

I'm going through a self-paced AI intro course right now that was put up online from Carnegie Mellon University, and one of the more interesting things the professor says right up front is that while they provide a validation framework to test your homework against and they hold back some pieces of it so you have a hidden checker... There is no ethics expectation that you won't use AI on the problem and you can submit the results against the checker as often as you want.

His thinking is quite simple on this... There's absolutely nothing he could do, practically, to prevent you from feeding the questions straight to an AI and having it kick out the answers. But since this is a self-taught, self-selected course... His advice is to not do that until you get well and truly stuck, and only then will you have enough attempts built up to compare to the answer to understand the concept you're missing.

I think it's a decent middle ground for an online published self taught course. But I don't know that such an approach works in terms of, say, a Carnegie Mellon University being able to continue to charge the tuition it does for the courses that aren't self-taught.

This motivation is why, when I was in grad school, I never liked handing out grades. I would have rather just marked the problems I found with a HW submission, added notes where appropriate, and that's it. If no grade is given, it is obvious that you have provided the student the service of reading and checking their work, rather than allowed them to increase their 'score' or collect 'points'.

> You have to figure out a new transactional structure that can’t have your tuition made irrelevant by a $20 subscription.

Maybe a problem in the long term, but irrelevant here. These students paid (a lot of) money for this service. The provider of the service needs to stop complaining and do their job (grade the homework). If even a single student did the work themselves, anything otherwise would be a disservice.

I'm doing law school through an unconventional program called LPAB. You have to complete 20 subjects, each with a 20% assessment and an 80% exam, with only a 3 hr lecture every week and that's it. You don't really get a chance to ask questions and there's no tutorials or handholding.

While the assessment can be gamed, you must overcome the exam which are written and reviewed outside of the university, so the lecturer doesn't know what is on the exam, and the questions and exam papers are never returned.

Doing this I remember my undergrad and masters felt like I was coddled as there was a bunch of fluff grades in doing assignments and just turning up to class. The grade inflation was a farce as well.

As other commenters said, graders are certainly not paid from grant money at US institutions. NSF (which funds most mathematicians in the US) would never allow it, and grants are usually much better spent getting grad students off teaching so they can get research done. Teaching assistant (TA) funding would ultimately come from the university. Sometimes it is optional, i.e, a faculty member can request a TA or grader for certain courses but does not get one automatically. Some courses, like large entry-level calculus, cannot be run without teams of TAs teaching recitations.

I'm on the math faculty at a R1 (public) university, and we've not had a budget for graders for most courses for a very long time, with the exception of e.g., very large courses. For most classes the instructor does all the grading. Grading has been a significant fraction of my time teaching, certainly much more than what I can spend on preparing lectures and other teaching-related work. Moreover, much of the time grading is giving feedback, figuring out partial credit, etc; just checking "correct" or "incorrect" is often relatively quick.

In the last couple years, my personal experience has been that (i) a very small (and possibly vanishing) fraction of students are declining to use AI, or at least use it sparingly, because they want to learn; (ii) a growing percentage are using it as a personal tutor; and (iii) some are outright outsourcing all the work. The last category is small, but if that category grows (anecdotally this seems to be the case -- I've mostly taught grad courses in the last year and don't have first-hand knowledge) then what would be the point of spending all that time grading and giving detailed written feedback?

Incidentally, many students in the second category are the type who in years past would form study groups, come to office hours, etc.; now they seem to prefer AI over humans. Some of them do seem to benefit from it, but not all. How one uses AI matters a lot, and how to help students use the tech to learn is something we have to figure out (regardless of how we feel about AI's impact on research).

There is no “earning potential” to be gained in a world where you simply ask machine to do X and it does it for you.

We might not quite be there but we’re trucking towards that point at full speed.

> They are going to college to increase their earning potential, and that’s about it

If we accept the narrative that LLMs can defeat the incentive structure of course work as a concept, it raises the question of if college will continue to raise earning potential for much longer.

Grading homework is still less degrading then the other shenanigans that professors use students for- as in ghost writting books, having them program apps as unpaid labour or doing similar stuff - including renting them out to other companies on campus as free labour. Academia is just very casty at times..

Some of my engineering professors had a policy that you couldn't get a lower grade in the class than your grade on the final. You could and they said some students did turn in no work at all and still got an A based on the final.

Graduate students spend 10-20 hours a week grading homework, it's a very time intensive process. But I agree, HW should only be an experience. The entire grade should be determined by an in class midterm and final.

> probably hasn’t graded a homework set in a while

Honestly he should just tell student to grade their homework with chatGPT themselves.

What’s happening to the signal?

That’s the goose that’s laying the golden egg.

Are the people that do well on the proctored assignments firewalled off from AI going to be the best employees?

In interviews I sort of bit the bullet and do pair ai coding interviews - you just have to change with the times

My guess based on my experience as a math major is that they just stop giving homework and the in class tests become more rigorous (can’t imagine how much harder they could have made these at my school at least)

My point is that no matter what tool you’re using, at some point you have to stretch your brain and the exams/screenings/etc just haven’t caught up to how to stretch someone’s brain in this new world and I’ve found just embracing the thing to be a better approach than any sort of pearl clutching

The irony is that homework barely helped even before AI. I feel like we should move more into self grading extra material for students to peruse if they are unsure about their ability in the subject, rather than adding chore to those that are fine and a point of worry for those that need that bit extra.

> You have to figure out a new transactional structure that can’t have your tuition made irrelevant by a $20 subscription.

No structure needed, the transaction was always and still is the promise of the social experience and connections for young people. It’s the actual business model; undergraduate education is break even, the only way money is ever made is alums donating to the school due to the feel-good memories they formed when they attended.

Unfortunately, for schools who thought their purpose was to educate, they will be made obsolete in the near future. Expect a lot of schools to close.

But those places that always understood their purpose was for affluent young adults to mingle and connect with other affluent people, they will be just fine.

> He has a youtube channel with lectures in number theory, category theory and other things.

The channel: https://www.youtube.com/@richarde.borcherds7998

Thanks for mentioning this. I just realized that the Monster Group has approximately as many elements as the Earth has atoms (|MG| ~ 8*10^53)

[flagged]

Students who need that kind of "stick" to do well on exams aren't college material in the first place. We have pushed far too many youths into college as the default option, and that hasn't been good for the students or the institutions.

There are plenty of people who have the innate intelligence and intellectual curiosity to succeed in college, but come out of high school lacking discipline and clueless about what they want to study. They would be better off spending a few years working or serving in the military or something to get their shit together first, and then enroll in college later. Those students tend to have clearer goals and take the educational experience more seriously.

When I was at university, a student caught cheating was cause for being expelled.

I am recruiting college kids right now, impeccable pedigrees, elite universities. Our recruitment process involves the kids recording videos where they answer questions as a pre-screening. They are explicitly told not to use AI for the exercise. I have the Claude and ChatGPT answer to those questions on my screen while I review their responses. 40% read it pretty much verbatim. The paradox is I don't even care whether they knew the answer or not, but to me it's an integrity test.

When I was young, public universities in Spain only evaluated the final exam. That was not a problem; it helped people become responsible. Nowadays, they have changed it to evaluate continuous homework, and students end up worse prepared.

Even ignoring learning outcomes, answer sharing was a good excuse for the nerds to talk to the not-nerds.

Back when I was a student, 25+ years ago, many professors were a bit lazy. They just re-used problem sets which they had made earlier. No single year would be alike, but the problem sets would still be from a larger collection.

The students would then hand down the big book of problem and solution sets down to the freshman class.

The professors, still being a bit lazy, knew this. But instead of scarping the old problems, they'd just say that cheating only hurts the students doing it - as the homework grades would not influence the final grade. And sure enough, the students that never did their own work, rarely did well (if at all passed) on the finals exam.

It should be social and other soft pressure to do homework, only the final exam should be graded.

Cheating always has been available as an option. AI just makes it faster. At Cal, you always had students who would still pursue doing things on their own even when not cheating led to lower grades. The reality of it all is that cheating catches up to you eventually and the later it does, the worse are the outcomes. If you cruised through high school by cheating, you’ll not survive college. If you cruised through college by cheating you’ll not survive your first job. If you cruised through your first job by cheating you’re not getting a promotion or a second one.

Half the class was cheating in Ion Stoica’s OS class even when the whole thing was supposed to be a group effort. Like how are you cheating when you’re already working in groups of 4?

The bigger issue here is that even if they don’t cheat and work hard to get all As, there may be no career for them at the end of it.

Even as technology has destroyed education as we know it, you and likely many others in education are still thinking about sticks?

The students have made it extremely clear to the teachers

[dead]

No that will not happen as the company has to pay and it shows in their accounts.

Much cheaper not to train anyone and let the state or pupil pay for training.

See lack of apprenticeships for a case of this.

I've heard this argument over decades and to this day it doesn't make any sense to me. You know what happens when you do that ivory tower thing? Those professions die.

I don't understand where the premise comes from that higher education is doing this, or even has it as a goal.

To the extent it is I agree it should obviously be replaced by on the job training, because that will be much much better. But to me this is what's happening now anyway, de facto.

Even you do, say, a computer science degree which is one of the more obviously industry aligned fields you're still not really learning how to build software products or even useful software at all really except in the most mechanical sense.

You're learning a bunch of variously useful theory (in the sense that it could all be useful in principle but you won't know exactly for what and at what time until you're out building things in the world) and no company is ever going to be incentivised to introduce you this in a proper environment conducive to intellectual exploration separate from 'the job', because frankly you'll be long long gone by the time any of it pays off often decades hence.

Maybe AI will be yet another disincentivized for higher education. When I was considering law school a decade ago, the overwhelming advices from existing attorneys was “unless you’re in the top 10% of the top 10% of universities, you’re wasting your time getting a law degree”. Seems very similar to university today (at least in the USA)

it is too challenging today to capture the value of training your early employees, labor markets too liquid

> With some exceptions I think higher education should probably go back to being about academia and not necessarily teaching skills for industry. I think companies can now easily invest in their own extended training and grab people out of high school. Just a thought.

They'll probably just shut it all down.

With AI, what's the ROI on educating the masses? Paying specialist academics to know things? Just leave them dumb and let save the rich some tax money. Dumb people are easier to control anyway.

On research: If some tech bro can't do the research in an afternoon with an agent, it's not worth doing at all, let alone worth paying a PhD for a tenured lifetime job to do.

Cambridge engineering had something similar called "standard credit" for minor assignments. You'd get the full mark by default, only losing marks if missing or egregiously bad. The exam at the end of each year carried almost all of the grade. It meant you had to engage with the work, but didn't turn it into an exhausting competition for the best coursework grades.

When I worked as a teaching assistant at my university, we had a mix of knowledge-based questions to which you had to research the correct answers, as well as some questions you had to brainstorm and argue your decision. We didn't grade correct answers, just looked over them to see if the student engaged with the question. In my opinion, those were really easy points to get, especially considering it was one of the hardest courses in the whole study program, but there still were many students who skipped the homework. Now, I've never been able to prove causation (wouldn't even know how), but it turns out that students who solved the homework questions also did well on the exams, and those who seemed to have used some AI chat for the questions did not. And even me telling my students exactly this did not affect the outcome of semester after semester. I believe that lots of guys were just there for the degree and looking to coast by with minimal effort, but for me it always seemed like a big waste of time to go to a university if you're not even going to try to learn something. The first few semesters, I was really frustrated avout this, but as the old saying goes: You can lead a horse to water, but you cannot make it drink.

Yes I also had this in Uni at some point. Just going through the homework can be a huge effort. If the course is small enough, you probably don't want to hand in obvious junk/slop, especially if you're asked to present "your" solution.

To provide feedback on mistakes and help them learn. Unfortunately if you use the grade to contribute to their overall marks, some will cheat.

But that's easily solved - just don't reward cheating on homework.

I agree, and most of my prof friends in math have switched from graded homework to weekly small quizzes. Graded homework was a way to use the stick as motivation because we've given up on the carrot (the learning of the material itself as an internal reward for completing homework).

Some of them keep the homework grade to like ~5% and mostly grade on completion because, when the student's homework proofs use advanced concepts and proof techniques not covered yet but their quizzes get sub-50 point F grades with barely coherent thoughts, it's a clear sign that they are not actually spending time using the resources the class has given them to work on the material.

It seems like all of them have given up on punishing students for it, largely because they're pretenure and the teaching part of tenure review here is primarily student evals drive (HINT: this is a MAJOR part of many broader problems). But to me it's kinda nuts that in one year or so, we have decided that it's actually ok to admit students into very competitive elite institutions, catch them passing off work that isn't theirs as theirs, and not treating that the exact same as we would have a few years ago if we caught them all sharing solutions.

Agreed. It's interesting to me to read about math homework being graded, because in my college math courses (at RPI in the late-aughts), homework was optional. It was "assigned" for students who needed more practice, but not required for those who didn't. It had no bearing on your grades at all.

I distinctly remember for the calculus courses, they told you a set of textbooks were required. By the end of the semester, I hadn't even peeled off the shrink wrap. The professor only used the books to cherry-pick homework problems from, and I personally never felt the need to complete them. I get that everyone learns differently, but making everyone buy that book never really sat well with me.

We did not have homework per se in what we studied but we did have reading that was relevant for the purposes of tutorials and being prepared for lectures. We also had deliverables (assignments, presentations, practicals) that had grading, and proctored exams.

I found having deliverables helpful in knowing how one is tracking with progress and understanding of the subjects and largely reflective of the final exam result.

I've also had studies done that exams were either 80 or 100% of the grading and that was a lot more stressful without any significant added benefit. In fact given that there were fewer or no deliverables, perhaps even worse mastery / retention after exam completion.

Homework grades are supposed to provide feedback and some incentive to do the work.

However, counting homework grades toward course credit also creates an incentive to cheat. It probably makes sense to remove (or at least drastically reduce) that incentive.

> If you don't do homework, it is only on you.

The issue with this policy is hyperbolic discounting — students are fine choosing not to do homework when the consequences (EDIT: that is, doing poorly on the exam) are still in the future, but less so when they become present.

Then they lash out at the instructor, even if it’s really their fault.

In high school math ,we usually had a stanupin the beginning of every lesson. Basically th entire class on their feet answering rapid fire questions from the teacher , usually some stuff that had been taught in the previous lesson. No calculator/notes just doing math in your head to get to sit down and start the class. Gnereally sucked when we were going through it , looking back though made digesting the material easier as you had to know, not just cram or hope for a calc or scrap paper to caluclate with.

So basically a oral pop quiz for math. i think could work well to complement homeworks by making sure the material as taught is actually inside the students heads.

That is not the point of the article, though. He posits AI will just do better than humans soon, and he no longer knows whether people should even study math (from the article). So the point is that pursuing those studies is no longer useful; how or whether you grade is irrelevant at that point. (To be clear, I am not defending the statement, just explaining what I understood.)

this is what I had to do in Italian university and we often had to do both written AND oral exams.

I'm willing to give him (or anyone else going through this) a lot of grace here as I think almost everyone who will be impacted by AI will go through 3 major stages, at different rates with each step taking a while to both really be true, and then a bit longer to accept and solidify into reality. And our egos act as blinders on accepting that each next step after 1 is likely.

1) AI is useless.

2) AI helps me.

3) AI replaced me (unless I am a generationally gifted top performer in my field).

While he is, in fact, a generationally gifted top performer in his field, he knows that statistically speaking most of his students will not be which I think makes his apparent angst now understandable.

The 100% exam weighting certainly makes sense for mathematics, but exam proctoring will have to keep up with technology.

but in the end what is the point of learning when you have a machine with all the answers and can do your work?

I heartily disagree, grading hand-written homework serves as an early warning signal to students who don’t understand a concept* but want to. It’s also useful to keeners who like being told "you’re doing a good job, keep it up!" As well a warning to students who are so far behind that they can save time money and their GPA by dropping the class early.

And as a TA, I liked the income, and there’s a knock-on benefit that you’re learning to read with a critical eye which will help as a reviewer in the future.

* or in mine own obsolescent field, a measure of accuracy that indicates a student need think** longer before writing in the future

** I know, I know

> Grading homework has never made sense anyway.

This is only true for subjects where there is only one answer to a question.

Plagiarism in other subjects tends to show itself. But AI has made plagiarism more viable in all sorts of fields.

I was an adjunct math professor at a University for a few years and I completely agree with this. There is this propaganda circulating in (and out of) education that testing is not a good measure of mastery, but it 100% is. It's a hard sell for public schools because, as you said, people will fail more.

Math is an even tougher sell because it's largely a service department. It's hard to justify holding non-majors up because they can't pass calc 2. For upper division math though tests are the way.

Yeah this is what confuses me and gives me hope. I would have loved to have had AI in college. I probably would have had the confidence to take more and harder classes and generally been much less stressed. 10 years ago it felt like most students would take any easy chance to cheat, especially on nightly homework freshman year. I wouldn’t do it because for some sick reason I was actually motivated by learning more than the grade.

I’ve loved using AI to learn. I made SQL and Python classes of my own to finally get beyond a basic understanding of them. I made an Anki deck to learn a foreign language with very poor existing online resources.

> On the flip side, AI's can now grade pure math problems with the accuracy and rigor of a postdoc.

Wait, I'm curious about this bit - I did my math undergrad ~16 years ago and all my homework was handwritten.

Are now AIs capable of ingesting my mess of poor handwriting with arrows indicating the flow of a proof (e.g. from the bottom of the page to the top of a hand-drawn column half-way up the page to cram in the rest of the proof)?

> Most of the times half of the class would get an F.

Doesn't that mean she's just not doing her job right? I mean, her job is to teach the class physics. If the class doesn't know the simple physics questions she's asking them, isn't that her fault? Why is she failing students when her teaching methods are so obviously not working?

Given stories like this, shouldn't we be glad that AI is completely disrupting our education system?

You could have said the same thing 20 years ago about wikipedia and the internet in general. A kid in a poor school on the other side of the world has access to the same wealth of knowledge than a student in a Ivy League university. But that hasn't made universities useless, and very few of those kids in poor schools (that wouldn't have gone to a elite university) picked up the opportunity.

If anything, AI in education I think will lead to a divergence. Mediocre students will be even more mediocre as they will use it as a substitute for learning, with AI doing their homework. Good students will leverage it to learn and understand things even faster.

> Having said that, higher education as we know it is facing extinction in less than a decade.

Sometimes I wonder what it must feel like to be able to so arrogantly make predictions about the future as if one had already lived in it. How can you be so confident about something you don't know for sure?

You're not wrong about grading being largely a numerical ranking system for resource allocation - with minimal effort from the allocator.

Even in the age of AI, I think I still prefer a portfolio system where you accumulate a body of work that you can present and defend in an interview for a job or graduate school position. Unfortunately diligent portfolio assessment can be a lot of work when there are hundreds of applicants for a single slot, so the first stage is usually reviewing a summary or list of work rather than the work itself, and that system can be gamed.

In any case, personalized feedback for student work is still very useful, and it may be more beneficial if it is focused on helping students individually rather than ranking them or sorting them into bins.

Without some sort of pre- and post-assessment, it can be hard to determine whether students actually learned anything in a course. The question is: how do you use that information? I think it should primarily be used as feedback to the students and the instructors, with the aim of improving learning.

So, the guy who does brain surgery on you... you want him to have taken some electives he was interested in, passed for having a pulse, then sent off to work in the operating room? Or, should he be required to have gotten at least 80% of questions right on a test of his surgical knowledge?

Well yeah, we just do in person assignments. We keep one every week, not necessarily a new topic every week, the difficulty is distributed, and take best K of N, which automatically allows for absentees and such. The system works well overall. For very weak students (bottom 10%ish) we customize it a bit and make a curriculum of sorts, sometimes re-covering content from a few weeks ago (AI is very useful for analysing answer transcripts and finding these weak points). And then, a final, also in person, for like 15% of the grade.

Large programming assignments also work decent (for now at least), where codex is encouraged, but we make the assignment research oriented and obscenely large with lots of messy and dependent steps, enough that you have to steer enough to do context management and not output slop reports/inferences. It is almost like a research project, where AI's "taste" is still extremely lacking (we often see totally circular/trivial-but-dressed-up investigative results from sloppers, very much like the slop neurips/iclr papers these days). Just one of these per course with a 10 minute in person viva at the end. Previously this used to be a project part of the course, but the old way with reports and presentations just ended up with way too much slop to use as a signal. Also, we just made it individual instead of a group project.

For this type of course, grading on a curve also made less sense, so we instead do a kind of "cluster" based grading. We start from the top giving S, see where the first big gap is, that starts A, the next big gap starts B, and so on. These clusters form really well in the above type of courses (empirically, undergrad lin alg, undergrad databases, and grad RL courses) and matches in-class judgement.

I'm a bit mixed on this. A lot of graded work is busywork, but a benchmark is needed to see whether concepts are landing or not before midterms or finals, and the easiest way for that is some kind of grading.

And why are you being downvoted to oblivion?!?

ChatGPT has been proficient in the English reading and writing department for even longer, and that shows in the quality of students. E.g. https://nypost.com/2026/01/23/us-news/gen-z-students-are-so-...

AI has replaced students in the same way as cars and elevators have replaced exercise

Maybe for "soft" subjects like History and Philosophy, where almost nobody who does those courses actually becomes a historian or a philosopher... But there are plenty of subjects where you actually do learn things that are useful beyond university. Engineering, law, medicine, chemistry, etc.

I had a class like this, Discrete Mathematics, with no grades or assigned homework and just a graded midterm and final. It was easily my least favorite class; I didn't know where to spend my time and spent most of the semester wandering aimlessly through the material. The final was very difficult but open book (to help make up for the lack of structure in the course?) and ended up having a ridiculous curve.

Could have also just been a bad professor though.

^^;

I have a solution, but I don't think people will appreciate it, because:

1) it is very analog 2) requires work from both student and teacher 3) there is no AI in it 4) there is not money to made on it 5) all of the above

In short, do what my MBA class did. At home you study the examples, the how behind it and possible solutions. In class, you may be called in front of the class to show that you know what you are doing.

Is it hard? Yeah, but it is kinda the point.

Today, your class would get shredded on student evaluations, the university would pressure you to lower the fail rate and if you resist some administrator would meddle and change things on your behalf. Ditching exams is a popular trend, for example. Not to mention that today your class probably comprises several hundred students at the small end and your student:staff ratio is like 40:1. Teaching today sucks.

I guess then we go back to the basics. The only constant is change.

Where are the successful alternatives?

It's already here. For instance I could walk into this guys math class, not even read the questions or material being taught and _guarantee_ that I'm going to be the highest performer as long as I have access to Astra/Fable. Like, it's not even close.

Even if you were right about AI, if you plan on keeping humans around, you need them as educated as humanly possible.

> university system (ignoring research and other functions, since that's mostly decoupled from the type of class in TFA)

The primary purposes of research universities is research. You can't ignore the primary purpose of the institution. To the extent teaching has fallen by the wayside it's because universities are research institutions first and teach and credential on the side.

Brian Harvey was really great as a professor. I loved his personality and demeanor that he brought to class. I came to cs61a for a SICP experience, and I'd say he delivered exceptionally well -- maybe more relaxed but equally engaging as lectures I've seen from Abelson and Sussman. It was very fun and rewarding; the course notes and assignments were great. To much regret, I only made it to his office hours a few times (was living in The City, and not an AM person), but I think he really cared for the students, and felt like a wise uncle that I wish I had.

bh's cs61a lectures from some years later:

https://archive.org/details/ucberkeley_webcast_itunesu_43829...