Even though it seems radical, I think the right approach is to simply allow the students to use AI to its full potential, to generate answers, code, whatever.
The onus should be on the instructor to make sure that the student ends up actually understanding and being able to code/solve problems that they pose without using coding agents.
Why? Because:
1. this is exactly what is going on in the real world. People are able to get AI to do whatever the hell they want, but the ones who just use it lazily end up with huge cognitive debts and codebases riddled with opaque bugs that they do not understand whatsoever. If we prevent students from confronting this temptation, then we are sort of coddling or shielding them from it, and not really preparing them to avoid pitfalls of this type.
2. you can actually learn a LOT by being given the answer, if you actually care to learn. i personally think it's pretty fucking lame to handicap a student's ability to learn in an attempt to prevent lazy abuse. isn't the whole point of a grade to measure how well you understand things? can't you have pop quizzes, assignments on a computer with no agent use, written tests, etc etc. to catch the lazy abusers? this is an unnecessary prevention of lazy abuse that unfairly handicaps learning
Stanford is a research university. The student should have full responsibility for learning outcomes. The university will provide support and opportunities to the extent its resources allow, but it's up to the student to choose if they want to take advantage of that. Those who need a more guided approach to learning can always go to teaching-oriented universities or find a personal tutor.
That's a major reason why employers have traditionally valued degrees from research universities, even if they are not particularly highly ranked. Being able to thrive in an environment like that shows a degree of independence and initiative.
> you can actually learn a LOT by being given the answer, if you actually care to learn.
Even if you "actually care to learn", this is a huge mental shortcut and you're deceiving yourself if you think deep learning is happening from looking at the answer.
On top of that, the pressures to just finish the coursework and move on to your other homework due tomorrow seems pretty high. Your suggestion means we're no longer coddling/shielding students, but we also aren't actively helping them, are we?
Not from simply looking at the answer. From knowing the answer and reverse-engineering or understanding how to arrive at that answer in the first place. It's not always the best way of learning, but it definitely is a great way to learn if you care to actually understand why it is the answer and how you would have arrived at it.
> Your suggestion means we're no longer coddling/shielding students, but we also aren't actively helping them, are we?
My suggestion is just the former, it doesn't imply the latter.
My understanding is that research shows more learning happens when the student has to struggle with the material to solve problems and answer questions.
I'm not talking about AI technology we use as the next generation of tooling. I am talking about new, possibly multiple intelligent species emerging which are superior to humanity.
i think we are all beating around the bush and not addressing the root issue: AI could obsolete human beings in the future. what i'm not seeing is discussion or exploration of different branches or paths that could occur once that happens.
I argue the opposite, that originality will actually become more valuable.
Think about it: everyone has characterized AI slop, as slop. Which means that we negatively value it in terms of originality. Combine that with the fact that there will be a lot of it, this means that original work will 1. stand out or be very distinct from slop, and 2. have its value amplified as a result of this polarization.
basically, we value originality more AND are able to identify it more readily.
related is also the fact that originality will literally be valuable as training data for future models
I can prod at a model as much as I want to produce something I find more original than average, but there are plenty of people out there that will say it doesn't count because of the fact an AI made it. "Slop" doesn't just mean "it sucks because it's bad", it often means "it's sucks because it's AI". They'd argue that if you were creative enough to produce something so original you wouldn't rely on an AI to make it for you. It's tainted by association, all the way back to the multi-billion-dollar enterprises that originally trained the models for their own ends.
Also there have been dozens of HN submissions and comments where the poster didn't even bother to remove the em dashes. Most people just don't care. The people who continue to post like this wouldn't have been as visible had they not discovered AI and pounced on it, but they were always there. The idea of posting with an AI voice, em-dashes and all, would likely have still appealed to them if you'd asked 5-10 years ago. Nowadays it takes hardly any energy for them to have a persistent voice.
I also define "slop" in a similar way. However, I specifically define it as creations that lack soul or originality. And can actually have a high degree of quality in some aspects, as you can see with some AI generated art and music. Because of this, I'm tempted to adopt a different term since "slop" feels too negative
Slop has always been around. AI has cheapened its creation.
Depends on how good the slop fork machine is, the act of true original creation is a messy and long process if it can be replicated to death immediately basically for free its not viable anymore
then it isn't a slop fork machine anymore is it? i was under the impression that the best the SFM would get is generating... how do I say this? high quality low quality work. Basically, the ability to cheaply produce quality work, characterized by its lack of soul/originality. think amazing looking advertisement graphics. not to say that it can't do better than that. just meant it as an extreme example for illustrative purposes
If something is able to generate things with soul and true originality... we're talking about something incredible, a new intelligent species potentially
It doesnt have to able to generate original things, its enough to be able to detect what makes it original and replicate the original thing with enough variations in different contexts to be able to be destructive and render the true original thing completely useless
think about how in music, when an artist comes out with something original and awesome, and then everyone starts copying it and creating their own derivative works, like Jimi Hendrix or something.
Did Hendrix become useless? Did everyone end up thinking he sucks or something? No, he is even more revered, as the originator of a new type of sound that probably created multiple genres
The same thing applies here. Originality will be valued and even empowered as extrapolation and development off of it can increase in speed and quality in the case you mention
> then it isn't a slop fork machine anymore is it?
True, but some nuance is that a LOT of artist/creative types lean exclusively on the mechanical skill needed to create, without anything really much to say. They also very frequently copy other's styles, etc.
I'm not defending AI pumping out crap, but this also shows a lot of folks don't have much to offer beyond the mechanical aspects and we shouldn't glorify churning out stuff by hand as high art either.
This honestly makes me feel ambivalent, as on the one hand, it is awesome that it is pushing creatives to be more original, but on the other hand it does threaten these types of creatives who have invested time into making this their livelihood :/
people need to reframe coding agent usage. i see a lot of framing in zero-sum terms where it's either all dev or all agent, and then people start dooming and glooming over the latter. in reality it's like that one post on here a few days ago about it being like an iron man suit. it is a glowing, bright white power that can be incredible when wielded properly. unfortunately, people characterize it as an adversarial power that can and will take over your soul.
how about some true synergy instead of boring zero-sum people? smh. the true poetry here is that zero-sum thinking will become more of a thing of the past so there is some natural comedy with this title
This is some anecdata, but I'll share it nonetheless as I have a pretty wide network of software and security engineer friends from which I've heard the following.
Almost no one I know wants agent usage to be a zero-sum activity. There are a few oddballs who obviously only got into software for the money, so any means to that end is acceptable. That does not stop those with say-so over things like employment (and, if you're in the USA, the associated healthcare), from treating it as a zero-sum activity.
When engineers are being told to maximize token usage, are constantly being brought into meetings where they're expected to reveal their latest and greatest use of LLMs, and not using enough tokens in your role is seen as a negative, then the pressure starts to creep in. Yes, I know this is silly to most people who read this site, and I agree. It's bonkers. But there is certainly something to the idea of "AI psychosis" in upper management that is making agent use zero-sum company-wide.
That's good to know, because all I really go off of is what I see posted online which is most likely skewed towards the polarizing takes (been unemployed for a while). Sounds like some positive news though, and I hope that these tools can help empower people to the point where they don't feel shackled by their jobs by doing something like lowering the barrier and manpower needed to succeed at entrepreneurship.
That shit with upper management sounds stupid af and I've heard the same type of shit from people I know who are in other fields. I'm guessing it's happening from a combination of ignorance, FOMO, investment, etc etc. But that is more of a systemic issue than anything to do with these tools imo.
I'm in a position right now where I'm trying to decide if staying in my own field of information security is worth it to me. I have an entire project plan built out for using local models to do some crazy augmentation of my own skill set, e.g. malware development pipelines and vulnerability research.
My biggest problem as an independent contractor is marketing and notoriety. Security has been a race to the bottom for over a decade now, but it's gotten exponentially worse. LLMs can't just do my job, but there are enough people with checkbooks who believe that it can and enough companies out there with an incentive to confirm that belief that it's getting harder for me to find work organically.
This psychosis you keep referring to will only end up punishing the ones who subscribe to it wouldn't it? Since it's out of touch with reality.
If that's indeed the case, then it sounds like an opportunity to get ahead of them since you know they will trip and fall at some point.
That's what I make of what you're describing, while sleep deprived and having given it some light thought LOL. So take that with a lot of salt. But it's kind of what I've been thinking these days anyways. Add to that that entrepreneurship is most likely getting empowered, and I think investing in yourself is the move these days. It will probably characterize the coming years strongly.
> This psychosis you keep referring to will only end up punishing the ones who subscribe to it wouldn't it? Since it's out of touch with reality.
It might if the subscribers realize/admit that things are failing because they were wrong from the beginning. But that doesn't seem to be how things work.
I would like to think they'll be the only ones punished, should punishment come. And as a disclaimer for what I'm about to say, I'm neither a Wall Street banker nor an AI company executive, so I don't want to accidentally make a specious connection between the two, but...
The 2008 housing crisis affected everyone. Bubbles that get too big pop across the population, whether they're complicit or not. As a little guy in a big world, with no expertise to truly know if there's a meaningful difference, I have a bit of anxiety about it all. I just don't want to catch collateral.
I just made a test with one of the AI:
It seems there is some evidence in there, sounds like mainly you are strengthening the muscles around your throat with that technique and this then can reduce sleap apnea a little bit.
It might be a purposefully sensationalist framing in order to increase KPIs. It works because a lot of people have strong opinions things without thinking much.
The onus should be on the instructor to make sure that the student ends up actually understanding and being able to code/solve problems that they pose without using coding agents.
Why? Because:
1. this is exactly what is going on in the real world. People are able to get AI to do whatever the hell they want, but the ones who just use it lazily end up with huge cognitive debts and codebases riddled with opaque bugs that they do not understand whatsoever. If we prevent students from confronting this temptation, then we are sort of coddling or shielding them from it, and not really preparing them to avoid pitfalls of this type.
2. you can actually learn a LOT by being given the answer, if you actually care to learn. i personally think it's pretty fucking lame to handicap a student's ability to learn in an attempt to prevent lazy abuse. isn't the whole point of a grade to measure how well you understand things? can't you have pop quizzes, assignments on a computer with no agent use, written tests, etc etc. to catch the lazy abusers? this is an unnecessary prevention of lazy abuse that unfairly handicaps learning
reply