You have probably heard the buzzword that has haunted many professions in the industry since the release of chaptGPT3, A.I. Every business wants to use it, but have no idea how they are going to use AI in their businesses. Every startup must have it to attract funding nowadays. Every business blogger has to mention it 28 times a month, for them to remain relevant. But do you know what no one tells you, creating an A.I. is almost the equivalent of building hell on earth, without the fire, fumes, and flames to make it work. It is freaking hard!!! Like really really hard!!!!
First, let's say you know how to program. Most people think, "I know how write a few lines of code, how hard can it be to build an AI?" Newsflash! So freaking hard that you would not even know where to start. Learning to program is just the first of the 101 steps that you need to take before you can even create your first model. Remember high school math, the classes that were so boring that you would give your left kidney if only you did not need to do a math test to go to college, well you should have paid attention in those classes if you had plans of creating an AI one day.
A.I. is simple terms, a freaking football field, in which all the grass blades on the ground are mathematical formulas and algebra. You heard me right. Let's say you did a computer science degree and paid attention in all math classes, ranging from probability for computer scientists to discrete mathematics. Well, for you, you are already 30 steps ahead of the majority of the common folk trying to get into AI. But you still have 71 more steps to go. Here is a simple list of the basic foundational mathematics that you need before you even peep into an A.I. classroom.
Calculus
Linear Algebra
Probability
Statistics
Trigonometry and Geometry
Discrete Mathematics
And these are just a few to get your feet wet so that you cannot be floating through the first classroom of A.I. But here is the thing, if you are anyone like me, you probably are good with basic arithmetic, but cannot remember jack about crap in any of the fields mentioned above. So, if you are trying to learn calculus, you will need to study first precalculus, then calculus 1, calculus 2, and eventually calculus 3, which is multivariate calculus used in creating AI. Now imagine that you do not remember any of the concepts mentioned above. You will need to study these topics from the bare bones to even have a simple idea of what the freaking lecturer is talking about when he says, "let's create a simple regression model, that uses gradient descent, in the next 30 minutes!" Because that is probably the simplest sentence that you will ever hear in an A.I. class.
But, wait, why all the mathematics? Why do we need math? Well, computers are made of 80% math and 20% software and hardware. Yeah, you heard me right. Computation is a mathematics concept. Computers compute. A.I. is like a third generation computer. Its a computer that does not solely rely on human orders to work, but one that thinks of what to do next. Its like the next advancement in computer programs. And A.I. works by prediction. Yes, you heard me right. Everything in AI is basically prediction. ChatGPT just predicts the next best word to fit the previous word it wrote. It does not work like you and me, where we come up with a concept in mind and look for the words to express the concept. No we are not even remotely close to getting to that level of advanced computation yet. It just spews a bunch of words, using complex formulas and previous training to predict the best next word to say after this one. And it works.
You almost think, "Woah! this thing can think better than I do." But it doesn't even think at all. It just get's your input- say a question you have asked, does a bunch of computations to look for the best first word to respond to you with, and then makes the best possible guess to say after that word until it makes a logical sentence that answers your question. Purely math. And then there is you, with a year or so of programming, thinking that you will break into A.I. within a few months, like the way you learn a new programming language or a new framework. Shocker!
You probably need an entire year of learning nothing but mathetmatics at least 4 hours a day for you to be a basic AI engineer. And Newsflash! Unlike the way you learned math in school, memorizing concepts and trying to recall the formulas in class or writing them underneath your shoes to cheat, you really need to understand the math. Like understand the math the same way you understand to speak in English.
The only upside is, the demand for AI products is just beginning. And with AI being as hard as it is, very few people will actually have the audacity to join the bandwagon. Don't listen to anyone out there saying that they can make an AI. If the know how to use frameworks like PyTorch but don't understand the math, they don't know jack about shit about A.I. Very few people will break into this industry and become seasoned experts. Are you going to be one of them? If you are, I will see you on the other side. Until then, Sayonara my people.