The codes of maths and man... Ramblings on the fear of lack of understanding in my studies of AI & ML.

Tis but another day and I continue my study of Artificial Intelligence, as my post the other day asking for book suggestions would itself suggest.

I searched and downloaded a heap of books. Though I have myself read some 20 odd books, or perhaps more, on python, there seem to have been many more published since I last went on a learning quest. Not that I have read no books in the last few odd years. Just not reading one every 2 or 3 days or possibly every week, depending on how dense the book and difficult to comprehend, as well as the time spent on it.

Perhaps I should have spent more time studying and coding...

But with these new books comes a fear.

I was never much for the code of mathematica, despite my learning. Not that I haven't studied the maths extensively. Every time though, it is like relearning and deciphering. Mathematica is a code that one must learn to decipher, preferably with ease. I'm better with programming codes.

For some reason, those that have taken over the fields of machine learning and AI have decided to teach it as a form of maths. Much of it is statistics and can easily be described using maths...but not all of it is. Often times they have to introduce ways to show concepts in equations that are foreign to most when it comes to mathematics...but are common in programming. Indeed, often times they are quite easier to understand in programming. Programming is more long form, with longer descriptive variables, and comments in the code to possibly explain it to anyone that would read. Equations are far more compact. They are good for writing on chalkboards and figuring out maths by hand, but not so good for deciphering.

Perhaps if I had a different life...different teachers...if I had never doubted my maths so...I would have not so much trouble deciphering equations...but if they were not so obsessed with mathematica as I am not...then I would have not so much trouble with subjects that I should understand more easily. Every day when programming I use complicated maths that I find easy enough, because I decipher and program them in code that I find easy and never turn to a whiteboard to write maths.

You don't really need to be that good with maths for most coding. You figure out a bit of math for certain things, copy bits of code from various sites, learn to patch things together, etc. Not that you "steal" code from sites. You look up things that you don't understand and use the techniques. You then make it your own, if you aren't an idiot. Though technically, you could probably patch quite a bit together copying bits from here and there. I'm sure there are programmers like that. Of course, you'll soon fall flat on your face, because often times documentation sucks and you have to just figure out things on your own. That's also part of why you often need to look things up and figure out how others have done it.

Because of the fact that so many that teach machine learning and AI do so with heavy maths, those in turn that learn it, for now, will be those strong in maths, and they to will teach it with maths most likely. It is sad, because it doesn't necessarily take someone strong in maths to code something, even something involving AI and machine learning. It may take a bit longer to understand something if one isn't so good in maths...but all the longer still when it's coded in something they have trouble with.

Machine learning is meant for machines, so why must it be taught in a language created for doing maths by hand? Why not in languages created for bridging the gap between man and machine?

The book I am currently trying to start says that many treat deep learning models as a black box. Indeed, it can be treated as a black box...but I have read many a book on AI and NLP and deep learning...and they're seldom as difficult as they first seem, once you understand them.

There is a common quote attributed to Einstein that if you can't explain something simply, you don't understand it enough. I don't know if he really said such, or something similar, and a quick search suggests that maybe he did and maybe he didn't but a few throughout history have said similar. Regardless, I've found that it may perhaps ring true. Many seemingly complex subjects are far less complex than they at first seem, once you understand them.

We have created many languages and codes over the years to explain concepts in the human mind. We have thousands of words by poets to explain feelings that are wordless in our own minds. Their words can convey emotion such that men are brought to tears that would not cry at their own father's funerals. And yet so many think that some concept they have in their mind must be stacked on a hill of education for a man to understand it.

Perhaps in some sense they may be right...but we are all learned men in this day and age. Many hold more books under their belts than scholars of old saw in their lifetimes. Much more can be taught than they likely realize, if but they considered more on how to teach it, rather than bringing the men up to the level to which they believe it needs to be taught to understand it.

I myself read quite a few things before I would be considered a man by my society that were at one time considered to be for those that had advanced degrees. I understood them well enough. I even read a few works by Einstein and understood them well enough. Though now I have likely forgotten most of it.

Once upon a day I listened to books by great minds in my sleep and argued with them in my dreams.

Am I to be treated like a fool that cannot understand the great minds that come up with simple equations to calculate the movings of the world because I have trouble deciphering their dense code?

Nay. The code they preach may be dense and cryptic, but it is no greater than the words of scholars and poets before them. The reality of AI and ML is that it is far simpler than we could have possibly imagined. That is how it is able to be computed by a computer so easily and simply that it can calculate thousands of records in minutes.

Of course, it's far slower for a man to calculate such things, because we don't have a direct line to the brain, and they often are not taught the tricks some know to calculate quickly. They have to decipher the code of maths, translate it to the brain, and then translate it back into the code. It's quite a feat that man is even capable of such a thing.

I will go to bed, and hopes that tomorrow I will once again fight to decipher the codes of man and read unto their books and learn of the teachings they would have me learn to teach things unto a computer...

_{Random mathematical formulæ illustrating the field of pure mathematics
Wallpoper
Public Domain (source)}