Yes, there is a degree of "mental muscle building" that a college degree is supposed to be evidence of.

The other issue is transferable skills. We know that someone who has become a strong expert in topic X is likely to have a much easier time becoming an expert in a related new topic Y. University degrees, even engineering degrees, are not mere "trade schools" where they just teach you the exact specific skills you need to get hired for a specific job title. They are teaching foundations, too, so the student can more easily teach themself new skills on the job.

If the student "successfully" avoids all drudgery in order to "focus on higher level thinking", they may well be avoiding building any true expertise in anything. And their "higher level skills" are arguably putting lipstick on guesses about things they do not understand.

Agreed.

When it comes to experts working in their area of expertise, I am fairly trusting that they will know how much and whether AI tools will improve productivity and/or quality of work. Experts have opinions and know how to back them up, or further investigate to prove out an idea an AI tool might offer them.

But what is true about people who are already experts may be the opposite of true for students. Students generally lack any semblance of expertise. They are at school to build up some significant degree of expertise in a field. Removing some amount of "drudgery" might have value, if it really means students are diving much deeper in other ways. The fundamental challenge in education is too much whittling down of the drudgery can leave the student's understanding completely untethered from the foundations of their field. They may not be exercising "higher level skills" at all, but putting lipstick on guesses based on feelings that cannot be supported by logical argument.

The related deeper issue people do not talk about is that a real education is about transferable skills. We recognize that having developed expertise in topic X usually gives a big leg up in learning a related topic Y. What if a student never really becomes a bit of expert at anything? And is using those AI tools actually a meaty important skill, or is it just going through the motions with software that will inevitably be obsolete in 3 years?

I agree.

It is like calculators. Calculators could potentially help students become more numerate than ever. In theory. In practice, for every student who experiments with their calculate to play around and explore arithmetic, explore polynomials by graphing them out, etc, there are a hundred students who are using calculators as a crutch to stay stupid. In reality, effective numeracy is growing worse than ever, businesses cannot find people who can run a cash register, and lots of people are too dumb to even get the calculators in their smart phones to help them when faced with even the smallest challenge.

Only AI is going to be 1000X as powerful in its effects. Yes, AI tools potentially could help students really master the material faster, get questions answered to become unstuck, do their homework more thoroughly, explore related side topics, generate additional practice quizzes, etc. In theory. In practice, for every student who really uses AI to strongly push their education forward, there are going to be a hundred to use it to avoid the hard thinking that is (was?) the point of university.

I believe it would genuinely to the students benefit to outright ban AI tools completely for K-12 plus the first two years of college. Yes, that is more draconian than absolutely ideal, but it is a draconian measure that is likely to be far better than the unfolding free for all.

What I expect will actually happen is in 5 years we will read in the news somewhere than 23% of the student body at Stanford has some kind of diagnosed mental disability for which smart glasses are part of the prescribed accommodation.

A flat universe is the logical default assumption, even if everyone understands it might be wrong. If you are going to posit a non-flat universe you have to come up with...something...to describe the non-flatness, such as Dark Energy.

No, these are not self-evident, you are correct about that. Unpacking the second assumption, what we now have to ask is what would a universe like that look like if the proposition is untrue?

We have two kinds of possibilities: (1) There is a unique frame of reference that is special and superior to all others, or (2) there is >1 special frame of reference that is superior to most frames of reference (but this becomes difficult to imagine how 2 or 3 or more especially special frames of reference do not end up contradicting each other).

Assuming the first is true, we would expect to see some kind of anisotropy in the universe that is otherwise inexplicable. In particular, we would expect the Michelson-Morley experiment to detect an "ether wind". Of course, that is not the only possible anisotropy, but it would have to be something of that ilk. Like particle physics experiments working slightly differently depending on the direction the accelerator is pointing at the time. (And, yes, particle physicists have looked for that.)

The race to build out more capacity also makes me wonder if there might be an intractable continuous obsolescence problem, i.e. 3 years from now a cheap knockoff will be 99% as good as the best we have today for 1% of the price.

With No Child Left Behind, there are funding consequences to test scores. The superintendent does not care if a student understands anything, s/he just cares about the test scores. The principal does not care if a student understands anything, s/he just cares about the test scores. The teachers do care about the students learning the material, but they have the principal breathing down their neck to improve the test scores. Guess what happens?

A neighbor of mine threw in the towel after 22 years as a kindergarten teacher. The principal was demanding what boiled down to test prep for the important 2nd grade round of testing, with the drills starting in kindergarten. My neighbor thought it was ineffective to the point of being unethical to drill 5 year olds for this bubble in test, when so many students were still mastering their letters, colors, how to write their name; at her request, she finished off her last 3 years before retirement as a 2nd grade teacher where she could wash her hands of the nonsense.

The problem was not the students or the teacher became dumber. The school system itself became dumber.

Now I am not so naive as to believe school systems did not already have significant problems. And I honestly believe George W Bush meant well, and, at the "powerpoint" view from100,000 feet up, the NCLB sounds reasonable enough. But where the rubber meets the road it has created more problems than it has solved. Emphasizing the material that is easy to test on a standard multiple choice bubble in test is not going to help most students understand the course material better.

Want to? No opinion. But it may actually be necessary to have something akin to emotions in order achieve an AGI that will be generally recognized as genuinely intelligent.

The problem with gold and silver is Russia is sitting on a lot of gold, and it is conceivable Putin will cash out a big market crushing amount to save his skin.

Not predicting it will definitely happen, but the risk of a huge drop in gold and silver prices is not so small.

If the junior developers are offloading a significant portion of the the hard problem solving work to the ML, they are simply less mentally prepared to learn from mistakes. There is less of a sense of ownership. There is less understanding of the situation that precipitated the mistake.

This comes down to how the human brain works and how it learns and improves at hard things.

Now, I am not going to assert that it is inevitable that an individual junior developer is going to learn less. I would say that using ML and pushing for apparent productivity is a situation that will very likely lead to junior developers learning VASTLY less.

Likewise, it is not inevitable that high school student using ML will learn little when creating an essay. However genuine scientific studies show most such students learn ZERO -- they cannot remember much of anything about the topic when quizzed after submitting their essay.

In a manner of speaking, my upper middle class in laws in NYC were very often "priced out" before this change. They owned a nice car and had plenty of money in their wallet, but the hassle of time in traffic and expense of parking was unattractive compared to taking public transport into Manhattan, most of the time.

If the non-rich now have a better run public transport system, the non-rich as a group are probably much better off overall. Having to go in extra early to work because the buses run behind schedule is a cost. Being late to dinner when one was expecting to spend the evening with the family is a cost. These things add up.

So, yes, the poor probably benefit, even if the positives are not necessarily a win for every individual, as is always the case.

One has to walk before one can try to run. One has to walk before one can try to dance.

This vision of producing capable independent thinkers who meld their own knowledge with that provided by machines tools sounds great. But how does a student get there, if they never ever stand alone and show they can think at all without the mental crutches? How do you judge their own capabilities without asking them to at least sometimes put down the crutches? How do they improve themselves in their role as AI partners when they have limited or no experience they can call their own?

At some point, a few colleges will throw down the gauntlet and ban all AI assistive tech from most classes, even if this means oral and in class exam are the entire grade. And what employers will discover is 90% of those graduates are pretty competent at many things, and 90% of the students from the AI lemming colleges suck.

To build on what you just said...

I am very wary of any argument that may boil down to "I do not not really understand what is going on here, but surely it will be easy to do better with a rebuild". Such claims are often founded on broad ignorance regarding the complexity of the real underlying requirements. And what I mean by "real underlying requirements" are those yardsticks by which the success of the project will be judged by the users/consumers who rely on the software. If the PRD was woefully incomplete you can argue failure is not your fault all you want, but you are still a failure.

However...something that is very messy primarily due to a poor choice of architecture, that is a reason to be auspicious about a rewrite being better. But do not be naive about the previous point. It is so easy to make incorrect assumptions about the requirements scope that your very promising new architectural approach will not save your project from disaster.

You handwaving argument that is so weak on specifics requires no refutation. The fact that you feel compelled to put words into my mouth in order to pretend to have a point is a mark against you.

You are right that "paradox" is not really the right word, but that is the label commonly slapped on.

It really a kind of Measurement Dysfunction, where an incorrect/misleading or fatally incomplete measurement drives counterproductive business behavior.

The underlying cause is the tendency to make linear extrapolations for all outstanding issues, where the poorly understood issues are presumed to be approximately as hard as well understood issues. Someone with a little self awareness would realize that if I understand X well and not Y, it could well be that is not random but because Y is much harder, the genuine difficulties hiding behind the cloud of my own ignorance.