"At this point I can only see self-discipline as a mechanism to willingly avoid AI"

Do you realise how difficult this actually is? Millions of people have zero self-discipline with their consumption of social media.

Yep! Which is exactly why I said implementing this at scale is TBD, I personally have no idea ;-)

One strategy I can think of is to remove the incentive to offload to AI, which probably means reducing the reliance on tests for evaluation. However, this also assumes intrinsic motivation within the students, which is also far from a given! Who's going to be motivated to do things the "hard way" when they don't even like the topic? Yet our curriculum is full of compulsory topics.

Like I said, none of our current systems are structured for this, and we'll probably have to figure things out from the ground up.

I learned Calculus despite having access to Mathematica, TI-89/92/CX-CAS, and WolframAlpha. I still had to do hundreds of derivations and integrations and manual manipulation of separable differential equations entirely by hand to learn it. But these tools made it easier for me to understand what I was doing wrong.

So, I agree with you, but it's also already been true for decades now with other tools.

All these tools only replace mechanical aspects, not thinking ones. AI is truly unique and unprecedented in that way.

A spellchecker is purely mechanical, it just helps you spell your essay right. But it won't make your essay good, or help you write the right essay.

Integration in calculus often required me to "guess" a strategy ahead of time. It was more similar to searching for moves in chess than solving long division. Some moves would make it easier vs. harder and some would seemingly be a dead end.

WolframAlpha usually got to the correct answer, but often used a non-human strategy for integration. Generally, the best way to do it was simpler than what WolframAlpha's "step-by-step" showed, which was rather "brute force" and inelegant.

So, again, I agree. But again, it's a matter of degree and encompassing more domains vs. a binary change.

It's a different degree sure, but it's a big enough leap I consider it a different thing. It's similar to paying someone to do your homework or write your essays. Your critical thinking approaches zero, which is really bad for education.

Not just education... it affects how people make decisions throughout life. In other words it makes a person ill-prepared to deal with the stuff that happens in life.

In another two examples:

As a kid, I remember my teachers not allowing us to use calculators and times table pencils exactly to make us go through the painful process of using our brain to get to know math.

Today I'm doing a Master's on AI and AI is a tool I use not only to summarize, but to validate my understading. I don't jump into a GPT session asking it to do the exercise for me or to explain something. I ask it to help me finding out my error after inumerous tries; to say the same thing me in another words because English isn't my first language and often I don't know the terminology that the Academia uses; to write an analogy (which often absurd ones pop up); to guide me where I'm missing something.

It's a freaking powerful learning tool if used properly.

Proud of my son as he's doing that without any prompting (ba dump tish). When I offered Claude pass, he said "I won't learn anything that way"

Your kid is gonna do well in life.

I think it all comes down to whether you want to learn something or have something. I can have an app in a few hours just like I can listen to a song on spotify. Or I can learn to code just like I can learn to play piano.

> The literature is clear that repeated, hands-on practice is really the only way to build skills.

The centuries of literature we have on this contrasts hands-on practice with theory: not actually doing the thing, but studying how other people do it, in order to gain knowledge that will be helpful when you get your hands into it.

This is different: this is like having a slave do it for you.

We know from history that the slave owners didn't know how to do the work. E.g. kings and feudal lords didn't know how to herd animals or raise grains, etc.

An apprenticeship is great for all sorts of reasons that AI can never touch, but I don't think abandoning AI will be necessary unless you aren't really motivated by a desire to understand and do the thing you are trying to learn. If you are, it is an incredible tool.

Here is a full proof plan to prevent students from abusing AI:

Pencil and paper exams, taken in person with all electronic devices prohibited.

Haven't been able to verify, but somebody on a podcast mentioned that sales of paper and stationery have started soaring precisely because of this.

Edited to add: I think this is worth trying, but not nearly sufficient. I would not be surprised if all we see is that grades plummet because until the pen-and-paper tests the students have just been winging it with AI.

Then any student who cares about his or her grades will change his or her study habits very quickly.

I would call that an optimistic take. It’s far more likely kids will complain, which leads to parents complaining, which leads to the teacher changing whatever was being complained about. From my experiences, school seems to not be a place where people take accountability, but instead, play a constant game of shifting blame when failure arises.

Yes, unfortunately, and we have already seen this happening for a while now with grade inflation:

https://www.ed.gov/about/homeroom-blog/addressing-grade-infl...

https://fordhaminstitute.org/national/commentary/grade-infla...

>"built the muscles"

You can build muscles by digging a field by hand but most people commercially use tractors. And operating tractors is a skill but a different one.

I imagine it'll go like that with AI. People will use it and learn the skill of using AI.

I mean you can learn the skill the machine does if you want like learning to write machine code rather than using a compiler but there may be a limited number of people doing that.

The problem with the tractor analogy is that there is not much that can go wrong in digging with a machine versus by hand. However, AI, like people, makes non-obvious mistakes that require a high level of skill to catch.

For instance, what we're seeing with pure vibe-coding is that even the best models make small mistakes, usually in the architecture but often in even lower-level abstractions, that compound over time resulting in unmanageable complexity. Senior engineers are able to spot and fix these issues early based on intuition, but that intuition has come from a lot of hands-on experience.

Using AI for a given skill is becoming a separate skill in itself, for sure. But this still requires having an understanding of the original skill itself.

I think this requirement won't change until we have super-intelligence, where AI is consistently much better than humans, i.e. beyond AGI. Because even at human-level performance, the only human that can verify an expert's work is another expert.

And the tractor operator with weak muscles is up the creek without a paddle in any situation that comes up that the tractor can't handle. They are also toast if the tractor's own advancements in automation renders less and less relevant.

Our only value in an LLM dominated world is the ability to handle the situations that the LLMs can't. Those who let their skills and critical thinking atrophy (or never built them in the first place), well, they're simply not going to make it. Shepherding LLMs is not a difficult skill to acquire and isn't going to be enough to let us maintain our value.

>You can build muscles by digging a field by hand but most people commercially use tractors.

Have you met a farmer? Most farmers are strong as all hell.

I also agree - however the ART of making software that's simple and works is disappearing fast.

One of the most impactful things in my career was watching Simple Made Easy by Rich Hickey

our industry is part engineering | part art (i.e having taste) - A.I doesn't have taste (hence why vibe coding) doesn't work - the art of engineering is reducing the amount of code you've to write

it's not about saying YES to every feature - but saying NO and shipping the most impactful feature in the smallest amount of time. sometimes it's about stepping back to relearn so you can do things the right way.

whereas AI is about producing the most amount of code possible - which is the very antithesis of Software Engineering

> the art of engineering is reducing the amount of code you've to write

I would argue that it's _scaling_ software development?

Making it so that you can create bigger programs: that accomodate more contributors, more requirements, faster, with more features, while being more correct, etc, without everything collapsing into spaghetti and chaos?

Historically, keeping a close eye on lines of code spent helped a lot along most axes. Finding ways to keep "code spent" down will remain very valuable but I don't think it's the _only_ way to scale?

Along the way we've developed architecture conventions, useful interfaces, test practices, standards, protocols, plugins & extension points, package repositories, etc. There are also social technologies like mailing lists, SIGs, bdfls, modern project & product management, etc etc.

To the extent that simplicity is measurable (less LoC spent, less code coupling, less cyclomatic complexity) it seems like it can be addressed as an optimisation problem and automated.

The literature is also clear: lack of practice leads to atrophy

Yeah I expect, just like physical exercise, people will need to get their hands dirty on a regular basis to stay competitive.

In fact, this is nothing new, and even encoded as a requirement for many professions like lawyers and doctors: to maintain their licenses, they are required to take hours of "professional development" every year or so to keep their skills up-to-date.

This process is regularly gamed, of course, but I suspect with AI people won't have that luxury.