Deep learning has been a buzzword lately. It is supposed to be something like machine learning, right? Is machine learning how you teach machines? How does machine learning work and how does deep learning work? Deep learning is really a subset of machine learning, so in order to understand deep learning, one must first understand machine learning.
The Inner Workings of Machine Learning
21st-century artificial intelligence is built on algorithms. These algorithms are the cornerstone of every computer model and computer program.
Algorithms can be improved through new data, known as training data. This training data is introduced to the algorithms through a cycle known as a training set. During the training set, algorithms are given a new task. Those that fail to perform the new tasks adequately are eliminated and those that perform well become the models for new algorithms. The standard for new tasks is raised and the process is repeated for another cycle. As the algorithms improve through each cycle, these cycles become known as layers.
There is a great number of layers in a neural network and they fall into many categories: First layer, second layer, input layer, output layer, next layer, hidden layer previous layer, and final layer. These layers are arranged in a hierarchy with an artificial neural network based on the biological neural network of neurons and nodes within the human brain.
Neural network architectures can vary from simple neural network, traditional neural network, convolutional neural network (CNN), and recurrent neural network (RNN). The type of neural network depends on the purpose of the artificial intelligence.
For example, a convolutional neural network is useful when managing a large amount of data within large data sets related to visual imagery and image classification, such as that used for image recognition, facial recognition and a sharper image with more pixels.
The Inner Workings of Deep Learning
What separates a deep learning system from a machine learning system is the concept of unsupervised learning. For regular machine learning, the training process is one of supervised learning, because the programmer has to personally administer each training set and learning set. This can take a long time on the part of the programmer.
Deep learning allows for unsupervised learning, algorithms training algorithms. This breakthrough allows the human programmer to spend less time since algorithms are faster than people. Because we can now accomplish more in the field of artificial intelligence, new fields have opened up: Data analytics, business analytics, predictive analytics, and data virtualization. It is a great time for data scientists!
A New World
Through the deep learning process, artificial intelligence has been taken to a new level and the possibilities are endless. Already we have new tools for security and law enforcement, the healthcare and health insurance industries, the investment sphere, and the internet. Will it be a bright future or a dystopian nightmare?
Elon Musk and other technology leaders have been asking this question. Some say that robots can replace people like did in science fiction works like R.U.R. and Terminator. Others feel that is the people that are controlling the robots that we should worry about.
Another concern is not the annihilation of our civilization, but the automation of our workforce that we should concern ourselves with. When the industrial revolution happened, shoemakers and other artisans were made obsolete. How many jobs are being made obsolete by breakthroughs in deep learning models as we speak? Will we all have to become data scientists in order to have jobs? What happens in a society where nobody needs a job?
As with any new technology, new knowledge does not come without a new price. Whenever our knowledge surpasses our wisdom, there is always a danger.