Machine Learning for Grandpas and Grandmas
Trying to understand technology for our grandmas or grandpas is very difficult, as much as trying to explain this to them. In this blog post we will try to explain what is machine learning as it we were explaining it to our dear relatives, and to do so we will try not go into technicalities, but we will go deep and will use simple terms so it is fully understood. Let us start by explaining what AI (artificial intelligence) is.
Artificial Intelligence and Machine Learning
Computers are very simple, you give them instructions and they follow them, they compute them, they solve them, they wont have to think, of analyze, or decide about anything. Then what is artificial intelligence?
The ability of a digital computer to perform tasks commonly associated with intelligent beings.
And what we mean with this is they will use data gave by us to learn, to make decisions, to plan, to understand languages, to solve problems, to recognize things like objects or sounds, like Siri or a self-driving car, they are receiving information and giving a result after “thinking” for the best answer, siri will hear your request and will look for the best answer that suits you, and the car will be aware if there is an object near it or the speed of itself and will correct the route or the acceleration. Let us remove the idea of a robot with human appearance also called androids, a very Hollywood and science fiction portrait, and imagine a machine that can play with you very well chess, like the first AI program written in 1951 by Christopher Strachey. It will try to beat the opponent. But how do the machines know what to do, here is where machine learning comes in.
Machine Learning is a field of computer science that often uses statistical techniques to give computers the ability to “learn” by training the algorithm who make this happen. And we train the algorithm by giving it lots and lots of data so it can adjust and improve itself. There are two methods for teaching machines to understand data, supervised and unsupervised learning.
But before explaining supervise and unsupervised learning, let us go back and divide Artificial Intelligence (AI) in two categories, general or strong and narrow or weak. Narrow or weak AI can do one of the activities previously mentioned and it will do it very good but it will lack from the others for example a machine that is very good at facial recognition but will do nothing more than this specific task, that is a narrow AI. Strong or general AI is a machine able to do all the activities mentioned before. No one has demonstrated anything close to Human level artificial intelligence yet, some argue is impossible.
Supervised Learning
The computer is provided with the output we are expecting, the purpose is for the algorithm to be able to “learn” by comparing its actual output with the “taught” to find errors, for example we give to the machine tagged images of cats and later it will be able to classify untagged photos of cats. It is used for example to find a model to predict stocks’ prices.
Unsupervised Learning
The computer is provided with no labeled data in order to allow it to find structure within its input data. The goal may be as straight forward as discover hidden patterns within the data set. For example we give the machine photos of cats and dogs and we let it find similarities between the different photos of cats, and the difference between the cats and the dogs, and it will finally be able to separate the cats and the dogs photos. It is used for identifying abnormalities in the data, that a human will not see.
But there is something very important in machine learning that we haven’t talked about yet, and it is deep learning.
Deep Learning
Machine Learning has many approaches, one of them is deep learning, it tries to recreate the human brain more specific our neurons and neural network. Deep learning is about layers in an artificial neural network (ANN), and the connection between them, the output of one layer is the input of the next one, like in our brain one neuron receives information form another neuron, it process those signals and pass it to the next one, and so on. Each layer in the ANN handle one task, for example the main goal is to identify what is on a photo, so the image goes into the ANN in pixels, and the first layer separate the pixel value and passes this to the next layer, the next one identify the edges, the next takes those edges and tries to make a form, the next takes that form and identifies the characteristics of it, and the last one says based on this last output what is on the picture.
In the artificial neural network the first layer is called input layer, the last one is called output layer and the ones in between are called hidden layers.
Basically what the Artificial neural network does is help the machine receive the information, split it so it can understand and process that information, and finally show the outcome.
The math in machine learning
Is it as simple as giving the computer data for it to learn? the answer, unfortunately, is no, it implies a lot of math.
Machine Learning theory is a field that intersects statistical, probabilistic, computer science and algorithmic aspects arising from learning iteratively from data and finding hidden insights which can be used to build intelligent applications. Therefor a lot of algebra, statistics and math are needed.
In Machine learning linear Algebra comes everywhere factorization, Symmetric Matrices, Matrix Operations, Vector Spaces and Norms are needed for understanding the optimization methods used.
Mathematics in machine learning are very important for some reasons such as selecting the right algorithm has influence for example in the accuracy or the training time, or the model complexity.
You will be thinking but why do we need to understand all this?, or why is Artificial intelligence or machine learning important? stick with us for a little bit more and we will see why and how Artificial intelligence and machine learning are important, and what do we use them for?
Importance of Machine Learning and Artificial Intelligence
If you have a smartphone or if you use an email, you have probably be helped, or be in contact with Artificial intelligence, your email classifies the spam without you saying an email belongs to that category, or if you constantly buy things in Amazon, it will suggest things to you based on your previous purchases or your search history.
Lots of business are interest on Artificial intelligence they want to automate some tasks usually done by humans. They could use bots in the customer service area actually there are many companies using chat bots who are learning how to respond in an intelligent way to costumers.
Artificial Intelligence automates repetitive learning and search in data. Instead of a human manually teaching a machine it can learn by itself with the use of machine learning.
Thanks to machine learning artificial intelligence adapts with the use of learning algorithms. Just like the algorithm can teach itself how to play chess, it can teach itself what movie or series to recommend next online.
We all know machines are faster than humans so they will analyze more and deeper data using neural networks it could save big amount of time in risky or crucial tasks.
Artificial Intelligence improves existing products by adding intelligence (excuse me for being redundant) to them, like Siri was added to the new generations of Apple phones.
Artificial Intelligence accomplish incredible accuracy. The more you use Alexa, or Siri or Google Search, they will keep refining the seeking parameter and you will get to the answer you are looking for faster. We could train machines to classify cancer at a better speed than we humans do.
And if the artificial intelligence was taught with unsupervised learning it would get more from the data than a human would and even more if you have the best data.
As we can see we use artificial intelligence in our daily basis, from our email, to our smartphone, even if we just watch sports, cameras have Artificial intelligence so in a game you do not missed anything important, or Netflix will suggest movies or series more suitable to your taste. We can dream bigger, imagine in manufacturing high-quality improved thanks to AI watching the process every time.
The sky is the limit with Artificial intelligence because people is talking more and more about it, and is has taken a lot of resources to be develop, think about it the self driver car is right on the corner, we have machines that translate in an instance. We could have an AI who identifies fake news at an instance and remove them from the Internet.
Ask yourself this do we have to be afraid of Artificial intelligence know that we know more about it, and we have seen its benefits?
