This post is an introduction to basic concepts of AI, a technology with an ever increasing number of applications.
Specific and general AI
Specific AI is highly specialized and trained to execute specific tasks. While the general AI is the one with intelligence level equal of human and self-conscious. Until now, it only exists on theory.
Machine learning
This is an AI field where machines learn to make tasks with algorithms called neural networks. From received data, they learn to identify patterns, make predictions with another data set, take decisions and learn with experience.
On machine learning, exist three types of learning:
- Reinforcement learning: this is the learning for try and error, interacting with the environment. The actions executed by the machine result in rewards or punishments, therefore, the algorithm makes actions to obtain maximum of rewards and minimum punishment. Some games apply this method.
- Supervised by humans: specialists label and structure a data set to train a neural network, so the latter can obtain data features to make the correct label, when it receives a new data set. Fraud detection and speech and writing recognition are some applications.
- Non supervised: the algorithm receives non labeled data and seek to identify features and patterns without human intervention. An example is the algorithm from news ou video site which seeks to know the user’s interest, to send recommendations.
Deep learning
This is a branch of machine learning, consists in use a neural network with many hidden layers, between input (red) and output (blue) layers. The name deep learning is because there are many hidden layers linked with each other, to process and analyze data. In addition to that, deep learning is non supervised.
