Neural Networks Taken Apart: Why our brains ??

So, with all the math advice aside. Let's dive into the real deal! Here's some things that I'd like to put out 1st, which will help you wrap your head around why we approach ML the way we do, and where the inspiration is from.

When you see the name 'Neural Network', you kinda' get the idea, it's based off of a network of neurons. You know the stuff we call grey matter. So if neurons are what make us smart, and help us learn, then it'd help machines too right ? Well that was the whole idea behind neural networks.

But why ? I hear you asking. "Humans are so poor at arithmetic! How does this solve anything ?"

Indeed we're bad at math, but one thing we're good at is learning.

But wait, I hear another question. "Don't computers have enough memory to store almost anything". Valid point. But that doesn't exactly help the cause either. Sure we can load all the information available on the internet into some data center and call it a day. But is that really the solution ? I mean isn't it the same as making the machine memorize everything. For anyone who's tried that for a math exam would probably realize how bad it's going to be.

So at this point we've a multi billion dollar computer that can give you pages of information about anything, all while consuming enough energy to run the planet probably. So now, if I asked it about the significance of Pythagoras theorem, it'd be clueless, as all it knows about the theorem is that it's kinda' like this..

Well sure it can use the formula to solve all your problems, but what about word problems ? Here's one from Khan Academy

Now, for all the hackers out there, it's easy to hack together a script to solve this. But then what about another concept ? And is it really the machine doing the work it's meant to ? So at this point, you're the teacher who's breaking down this question and giving your student (the machine) just the numbers asking it to solve for the side length. I hope this clears up why memory is not the solution, and neither is pure arithmetic supremacy. You maybe able to perform the hardest of the hardest calculations, but what's the use if you can't understand the problem to get the data you need from it ? 

I'll see you in the next post, where I'll be comparing the human version and the machine version of a neural network.