"The process of simulating actions is the process that changes the structure of your mind"
0. https://www.sciencedirect.com/science/article/pii/S187705092...!
1. https://www.creativemachineslab.com/uploads/6/9/3/4/69340277...
* Childhood development. A child who has had early childhood adverse experiences mind is effected. A fMRI or chemical detection can see physical differences in the brain. But the brain is plastic (it can change) over time. Works on this subject have been published by Dr. Dan Siegel https://drdansiegel.com/books among others. The key insight here is that academic learning is not significantly different at a core aspect then behavioral learning.
* Physical sports/martial arts depend on a reaction time much smaller then what is afforded by going through the full frontal cortex. "muscle-memory" isn't real (it isn't "memory" as you think of it). What you have in these cases are "short-circuits" (this implies structural changes) that are able to act before you are consciously aware of what is going on. The same applies to math facts and other fundamentals, you move things away from memory that needs to be retrieved and into reaction. Reading C-syntax for programmers or signing your name is something that has been turned into structure that doesn't need "memory".
I think your initial premise is correct. We have limited memory. How do we overcome that? We write. Writing is important because with it we can overcome our natural memory limitations. You cannot think about complexities (well or clearly) if you cannot write.
The danger of writing is that you can produce something that is both irrational and nearly impenetrable to the casual reader. For example:
> Foucault's use of the concept is descriptive, that is, analytical and explanatory, and at the same time normative and critical: he describes the grip biopolitics have on individuals through technologies of power in a way that makes manifest the repression at work in these biopolitical processes.
The above, taken directly from an "academic" published journal, could be said to have meaning. Unfortunately, each one of these words has an alternate meaning that is not normative to English, making the entire (actual) meaning opaque. "contecpt" "analytical", "explanatory", "normative", "critical", "biopolitics", "individuals", "technologies", "power", "manifest", "repression", "processes" are all defined differently then a standard English dictionary. So even if you can get past the convoluted sentence structure, the intended meaning will still elude you.
But the answer to memory limitations is clear writing using common definitions of words. I bring this up because as you extend your memory beyond what it innately has, the more likely you are to fool yourself (and others) with sophistry.
Very interesting. Is the theory here that by writing things out on a page we are then able to manipulate the ideas in our head without the usual limits of our working memory? Working memory is still limited but because all the information is so nearby and within view we can quickly put things in and out of our working memory so its limit doesn't impede us as much?
But does it effect the field like programming? When programming, if I can remember the context, then I can easily search it (research paper, books, documentation, forums etc) Now, with Co-Pilot, isn't it effectively beneficial to understand a topic and develop a general problem solving framework for ourselves so that we can let the AI do it's thing?
What are your opinions regarding it?
Specifically you won't be able to solve a programming problem if the answer requires you combining over 4 things you don't have in your long-term memory (even if you can look them up). This is the main reason why Jeff Dean is a better programmer than me even though we both have access to google - he has more knowledge & experience of programming in his memory than me that means that even though we can both look things up he is able to solve way more problems than me.
Co-pilot slightly changes the type of thing that's valuable to remember, but it doesn't change the importance of remembering things. I think, as you implied, co-pilot probably makes remembering some types of syntax or boilerplate less important.
I think what the article misses is that there is a ton of knowledge we can't 'write down' and therefore memorization is not enough. For example learning to solve Integrals. Yes, there are some rules and tricks one can memorize which helps but I would argue the only way to get good at solving integrals is to interact, resp. solve them.
Another point are second order effects of how one learns, for example curiosity and resilience. There might be a long time negative effect on motivation of a topic, when there is too much focus on 'memorization' (It certainly was that for me in my French class;)).
I really enjoyed the 'bad reputation' part and agree that it is sadly viewed as not important enough by many.
Speaking completely personally, forgetting a lot of information is the easiest way to ruin my motivation to learn a topic.
[0] https://link.springer.com/article/10.1007/s10648-022-09677-2
Anyhow, you can learn to have better "interfaces" to your automatic unconscious abilities and leverage them to e.g. remember instantly and durably any mathematical equation, etc.
- - - -
As an aside, the strategy you sketched out at the start of your article would be workable if you have great recall, however an even more better strategy would be to recapitulate the discoveries of physics in roughly historical order. Your understanding of quantum mechanics would be much deeper and richer, and you would be following a story (an epic story made up of so many fascinating smaller stories, and one that is still going on! Albeit things have calmed down this last century or so, but no one thinks we have reached the climax yet.)
That would be the way to do it: start with the Greeks and the Alchemists and proceed to follow the trail(s) of how we as a species sussed out the mysteries of the physical Universe.
For example, in machine learning, there is something called stochastic gradient descent, where to learn you present a single random element at a time from the dataset. In the end it will have learned of all concepts, by becoming more and more confident in each individual concepts.
For example to learn QM, you pick a random QM wikipedia article, and try to push through the article, even though there are some things you don't understand. Then you do the same thing, for a different unrelated QM article.
For learning tennis, you don't learn specifically forehand, then learning backhand, but you alternate them at random so that you have a single unified way of playing with smooth transitions, instead of having to switch between different "modes" of thinking.
Sure more memory can allow some speed-space trade-off in learning ability, but using your memory too much may make you miss some fluency that may have emerged. For example the old-school of machine learning was using databases and K-near neighbors, which used a lot of memory and was slow. But the new-school of machine learning are using constant memory algorithm and compressing the data in it, and it can learn to generate all the pictures in the world with only 4 Gb of weights.
Learning is imagining, once you bootstrap your imagination, its bandwidth to synthesize new examples from which you can learn from, is much greater than the bandwidth of looking up new data material to learn from.
however long-term storage is just one vital factor, another one is the 'deep learning' neurons that understand the content it stores and more important to connect the dots among various neurons.
we need both: understand and store. Neurons do both for us.
My takeaway from this article is that if you only focus on understanding (and so do not commit it to memory), you cannot reason using this information in unfamiliar contexts later on, once you've forgotten it.
So the best thing to do is to: 1. Make sure you understand something thoroughly 2. Test yourself on it using spaced repetition to ensure you keep it forever
For transparency: I'm one of the cofounders of Save All (linked site) alongside Petros the author
I am utterly baffled that no response in this thread so far has taken issue with the statement "As you probably know intuitively, it won't work." For me, this _does_ work and I have proven it many many times over the years by adding entire categories of technical knowledge to my repertoire. And not superficially, either - I get paid very well to do things professionally that I taught myself by reading Wikipedia.
If my experience were commonplace, the "it won't work" statement would be highly contentious in the comments here. Since it isn't, I guess I can deduce that I must be an outlier.