Why AI?

​So... why should you care about AI? (Of course besides the movie generated fear that one day "robots" will take over humanity)​. Well, this is why I care (I'm sure I'll keep adding to the list but for now): 

1. Today a part of AI called machine learning (ML) is making many decisions that involve people, but it's used as a black box. ML can decide if you get a loan from a bank or determine your evaluation score if you are a teacher in Washington, D.C.
2. The results of something call Deep Learning are showing something similar to human creativity and in some cases have won contests of human creativity - So, what differentiate us from AI?
3. I have a daughter and I'm Mexican-American. So how are we going to help children and non-technical people understand this new transformation and participate in it, instead of suffer the by-products of it for decades and then try to change it after a few generations and lots of suffering (as we have historically done). 
4. We are creating a learning creature?!?!?! that can have as complex thoughts as us?!?!?!?!? What can this tell us about ourselves?​

Ok. I read Chapter 5 from the Deep Learning book. I learn a lot! But let's start with the basics: 

Humans have created different mechanisms for a machine to learn, most popular are: unsupervised learning UL, supervised learning SL and reinforcement learning RL.

Unsupervised Learning is used to learn a structure of a data set, and this is analogous of us finding patterns on a certain experience. We use this type of learning a lot, an example can be when we see a map of a new place we have never been in and we are trying to make sense of it. We can identify areas with a lot of restaurants, that are well connected through public transportation, areas with gardens, residential areas, etc. Nothing is explicitly labeling those areas for us but we use our pattern matching to learn about the layout of the map and find how the different features of it (green areas, restaurants, transportation) are distributed on the map and what patterns they form.

Supervised Learning learns features that are associated with a label, you might be very familiar with it. It's basically the favorite way many schools like to teach. For example: you may have to learn about the parts of the digestive system and learn to classify each of it by it's name (label). Later your teachers will give you a diagram of the digestive system without labels and ask you to correctly identify each part and label it. 

Finally Reinforcement Learning learns to maximize rewards in an specific environment. For example like with poker, someone can tell you the rules of it but you need to play it in order to experience winning and losing. Then you can actually learn what are the best strategies to maximize winning.