I’ve always spent a lot of time for nothing and while that is my burden to bear but I have yet always tried to be better. Here are the confessions of my self-betrayals.
Not putting in the the time
A very obvious one, if I don’t put in the time, nothing is possible.
Not putting in the “effort” in that time
Putting in the time gave me the “satisfaction” that I’m doing something but often I was just doing it passively, without concrete goals or notes to go with my learning. I was going through it like how an average student goes through a class, sitting there, listening, sometimes thinking but not really “engaging” with it.
If you go about the task “consciously” and in a structured manner, you will get much more out it in the same time.
To tackle the above points individually:
What is a goal/plan
A goal means a lot of things that just a vague “I want xyz”. It need to be as small to be immediately achievable within the day or any other smaller immediate time frame you have. Anything larger must be broken down to such goals. This is how you should start with it and is “popular” knowledge.
The above has a flaw though, that it requires you to have prior knowledge about the subject to make an effective goal, which brings us to a circular dependency making the whole thing redundant. Another flaw is that it does not allow for branches and deviations to learn more things as pre-requisites that are also not known prior.
My overall refined approach is to study like a dfs algorithm. I always have an overarching goal, and an immediate goal that is a step towards that. As I start to work on the immediate goal, I have full freedom to freely change the immediate goal is some pre-requisite comes up ( so I cover the entire subtree and then return back ). There can also be cases where I encounter something new that is not a pre-requisite but something to cover later, this gets pushed to a stack that I must maintain on the side.
It is also valid to push exploring a node completely if enough information is learned to covert the current immediate goal. Though in this case I should push it to side stack. Since there is no “real” graph, you can argue that this is the same case as above.
Measuring time
As long as you are following the above, you don’t and should not measure time. It will only create a false sense of urgency leading to a more dangerous shallow knowledge. This also creates a sense of time breaks and disrupts the “continuum” that should exist.
Taking notes
This is my primary litmus test of whether I’m actively learning or not. If there’s no notes, that information is not well researched and prone to be forgotten in which case I might as well not do any of it in the first place. This is one my most common self-betrayals. Where time goes by and I have nothing to show for it. Having notes is also a mental way to “commit” that information in a visible form so I know I have reference to the things I have done, instead of not even knowing where the time went by.
Seeking fancy overhead
I should not try to improve the above ( simplification is a different thing ) by, say, adding a day or month end review of things etc, that is additional complexity. I do not want to spend my time and brain on needless things and anything that adds “more tasks” to the above is not needed.
Not all learning is equal
There are different way to go about it. What you do ends up defining how well and how fast I you end up learning.
To be better at X, just do more X.
is somewhat correct but then you see cases where different people, both of which put in the hours and end up with vastly different results. If someone is simply limited by their own intelligence then that’s a different issue though since this does not apply to me, I don’t care. This is not about that.
A better way is:
To be better at X, do more new X in a sane way.
The sane way here is my way as describe in the second point above. The “do” part does require a more thorough discussion though.
What does “do more new” mean?
“Do” can be context dependent but it’s simply to perform the actual task that you want to be good at. For math, do is solving more problems, for programming, it’s writing more code. More means “do” should take precedence over something that’s say, not writing code. This is to avoid the fallacy of spending too much time in theory without actually doing the thing to cement that knowledge. The “new” is to avoid the fallacy of contantly “doing” with no reading on new things, leading to being stuck with that same knowledge and not improving.
I like to think of it in machine learning terms, especially gradient descent. The “do” is a step in gradient descent, this is what ultimately takes you to a minima. “new” is the learning rate ( or temperature/step size in a way of thinking ). It not only allows you to “learn faster” ( since you are backed by theory and not just figuring things out by yourself ) but also makes you not be stuck in a local minima. “more” is simply how efficint or capable your system is at these tasks, is it a dedicated cuda cluster doing a lot of these steps in a finite time of just a single cpu barely managing a few epochs in that same unit of time.
( life is all algorithms )
The place of books and videos
If you read cover to cover, and just read or spend the majority of your time reading, it’s being inefficient. Videos are an even slower way to transfer information but they have their place for when you are not at your 100%. Doing things in the most optimal way incurs a huge mental cost and drains you very quickly. It’s not a worthless drain though, you end up learning much better and much faster than through a video. Though between just reading and just watching, I remember something longer through a video simply because the same point is being iterated multiple times and in much slower way.
Though considering I’m trying to be in the most optimal setting, books and videos are a great way to figure out what are the new unexplored nodes, but without being accompanied by the “do” part they are not very useful and only give a fake sense of progress.
The place of AI
With LLMs, I am of the opinion that they are the current best way to learn new things over books and videos. Maybe not the best to explore new nodes ( since they regurgitate popular common stuff ), but if you know the topic and are being “inquisitive” about it, the back and forth dialog that you can get whole “doing” things is priceless, but only if you use it akin to a very knowledgeable mentor. A mentor who can guide and answer your questions but not do the work for you, and the later part is critical.
But I feel tired…
Caffeine.