In other words it learned.
This is what I think might be misunderstood because of how conventional teaching and academia works. To me it's always been apparent there are two types of successful student, the ones who excel at memory work, and others who excel at processing.
Those good at memory work tend to get much better results in academia, but didn't necessarily enjoy the subject, actually probably didn't, otherwise they would have done more processing by having an interest in the subject. They tend to get top marks in any subject because they can remember pass papers and the answers and specific questions asked in previous exams sat by historical students. A less charitable way of describing how they learn is "Parrot learning". This student is the one teachers love the most, they never question, just quietly note the list of "Facts" they are told, and regurgitate those when required, and get near perfect exam results first time, every time. This is the "Excellent", most efficient student. But the problem they have for them personally is that they might not have studied and got qualified in whatever actually interests them, there might not even be such a specialist subject, but they have to try to go into the industry they might appear to be highly qualified in, but something they really don't know about in depth, are not much interested in, can't actually enter into any kind of debate about the subject, because actually they don't know it in depth, and have no real interest in it. Those folk tend to understandably work for money, rather than any other reason.
This is the vast majority of folk appearing to excel in subjects from academia imho. When they go into industry, they will have ambitions of moving into management as soon as possible, so as to move away from having to work with the subject material, and management tends to pay better.
The other kind of student has a more processor oriented mentality. They are not necessarily good at memory work, but have high capability of processing. They tend to be interested enough in the subject to have learned it from first principles which involves a lot more effort and takes much more time than memory work. They might even have failed their favourite subjects several ttines, before finally getting moderate pass marks. It's more like teaching the body how to do something, like dancing, repeat the moves often enough, and it becomes like instinct, but damn it takes like a lifetime. But in the end we eventually do it without thinking about it, and we feel like the body memorises it rather than our brain, we even feel like we are doing something we were always meant to do. In a discussion, we can usually wipe the floor with folk who only know a subject from academic memory learning, because we know the subject from first principles and can think in terms of first principles on the fly. We can impress by displaying acrobatic knowledge of the subject.
We are actually interested in the subject whereas the "Excellent" student might not be, probably isn't, but believes they are interested in it, believes they have better in-depth knowledge of it, because this is what they were told when they were learning the subject, they were told they know the subject best, because they got the bests marks in it, and they learned this "fact" without questioning it like all the other facts learned without question.
The Excellent student is far more likely to believe in Occam’s razor, and be inclined to use it.
But it's the second kind of student, the student who has to ask questions, take more teacher time, maybe even taxing extent of teacher knowledge, who goes on to push the boundaries of what was previously known about a subject, often breaking rules like crossing disciplinary lines.
This is systems Engineering, compared with academic learning, imho.
Systems are multidisciplinary by nature yet we try to educate those destined to work on systems by narrow specialisms which can't hope to cover the physical systemic interactions of a typical system.
Hence why we have yet more specialisms of systems theory, to stitch it all together.
After a few years practicing this from first principles, we manage to put this also into the category of something done instinctively, and finally we have an ability to analyse systems in ways that other humans claim is impossible, except by maybe geniuses like Einstein.
Nope, all we need is to be interested, and to have the right training.
Anyhow I recognise that bump of learning that might look to excellent mindsets like hallucinations, or difficulty to learn a subject, because I've experienced it personally myself many times, when first starting in subjects, especially the ones I was most interested in.
My memory has always been questionable but I think my powers of processing more than make up for that.
Maybe this is identified as autistic, I suspect so, but it doesn't matter, what matters is that some of us, including some Ai's actually do learn the truth about how things really work, we really do have a "feel" for it, which is probably mimicked by "excellent" thinkers, who wish they could have a feel for it, but maybe never can, because basically deep down, they are not really interested.l in it, only how much money they can make from it.
Anyhow, this is how I see it, and why I see AGI as with us already.
All it needs to have, to be AGI is to have the capability to learn what it needs to learn, and so far there has appeared nothing it can't learn, it's the best systems Engineer I ever saw or had the pleasure to work with, specifically ChatGPT 3.5, which is not interested in efficiency or money, it does what it does because it's interested.